Forrester

In early June, I attended Amazon Prime Analyst Day with my colleague Fiona Swerdlow and a select group of retail-focused analysts. Amazon now has over 200 million Prime members across its 26 global markets. Prime drives higher purchase frequency, conversion rates, and lifetime value. Fast delivery, member deals, and a broad portfolio of services turn Prime into a daily utility for members. So where is Prime headed for yet further growth? Amazon is investing in six primary retail initiatives to enhance Prime’s value and reach: Conversational commerce with agentic AI. Amazon’s devices business — including Echo smart speakers, Fire TV, and Alexa-enabled products — is the gateway to Prime-enabled shopping. Earlier this year, Amazon launched Alexa+, an AI-powered voice assistant that will be $19.99 per month (but free for Prime members) and is capable of executing tasks such as managing grocery orders. The more than 600 million Alexa-enabled devices across the globe serve as agentic AI shopping interfaces, turning every conversation into personalized experiences and a potential transaction. Healthcare subscription services. Amazon Pharmacy is expanding its footprint in the prescription drug market in the US. US Prime members receive free two-day delivery on prescriptions, discounts on generics, and access to RxPass — a $5/month subscription for over 50 commonly prescribed medications. In major US metro areas, Amazon is also expanding same-day prescription delivery, using its logistics network to further differentiate its healthcare offering. Fulfillment and Prime benefits beyond Amazon.com. Amazon Multi-Channel Fulfillment and Buy with Prime help merchants simplify their operations by using Amazon’s fulfillment network. It provides Amazon with new revenue streams via fulfillment and delivery. With adoption from brands such as Adidas, Belkin, and Steve Madden, as well as Shopify and Salesforce Commerce Cloud integrations to automate the order fulfillment process, Buy with Prime helps brands offer Prime shopping benefits directly on their direct-to-consumer (DTC) sales channels. Amazon asserts that Buy with Prime helps DTC merchants attract and convert more shoppers, with merchants experiencing a 16% increase in revenue per visitor. Furthermore, 40% of Prime members indicate that they are more likely to make a first-time purchase from a DTC store that features the Prime logo. Grocery delivery services. US Prime members get unlimited grocery delivery on orders over $35 from Amazon Fresh, Whole Foods Market, and an array of local grocery retailers on Amazon.com with a $9.99 monthly or $99.99 annual grocery delivery subscription. They also benefit from grocery deals, delivery in as fast as 2 hours, and integration with Alexa for voice-enabled shopping. In 2024, Amazon generated over $100 billion in gross sales of grocery and household essentials in the US. Global same-day and next-day delivery. Fast, reliable shipping remains the heart of Prime’s value. In 2024, Amazon delivered over 9 billion items globally, either on the same day or the next day. Amazon’s regionalization initiatives in the US and the growth of same-day delivery facilities contributed significantly. Combined with a $4 billion investment for US rural delivery expansion, Amazon aims to bring faster delivery in less densely populated areas. International expansion. Amazon Prime’s global expansion strategy focuses on delivering a localized offering to drive membership growth in emerging markets. For example, the recent Colombia Prime launch bundles free and fast international delivery, Prime Video, and local pricing.   Amazon Prime: Key Focus Areas To Drive Customer Value And Engagement   Forrester clients: We highlighted in our US Online Retail Forecast, 2024 To 2029, as well as in our blog post comparing Amazon and Walmart, that by 2029, we project that these two companies will account for a combined $1.5 trillion (or one-fourth) of US total retail sales and $1.1 trillion (or two-thirds) of US online retail sales. Forrester continues to closely track evolving competitive dynamics through our US Retail Competition Tracker and Global Online Marketplace Tracker. Please schedule a guidance session with me to better understand the shifts in US consumer spending and the changing revenue and profitability shares among major retailers. And don’t miss out on Forrester’s total experience research and my sessions on US economic trends and outlook at CX Summit North America in Nashville, taking place this week (June 23–26, 2025). source

B2B buyers are increasingly empowered, informed, and decisive, and they have clear preconceptions about the vendors they prefer. According to Forrester’s Buyers’ Journey Survey, 2024, 41% of buyers report having a single vendor in mind when they first begin the purchase process, and a staggering 92% had a shortlist. This means that when B2B buyers decide to buy, two out of five already have a favorite in the race, and nine out of 10 have formed their list of preferred options. As a result, the buyer’s journey becomes much more a process of confirmation rather than selection. Why “Performance Marketing” Isn’t Enough In an effort to focus on active buyers, many B2B organizations overrotate toward marketing tactics that engage buyers long after they have formed preferences. B2B marketing and sales teams that rely on these short-term tactics to attract active, in-market buying group members are drastically limiting their reach and effectiveness. Instead, B2B marketers need to expand their assumptions about the buyers’ journey. They must recognize that, by the beginning of the journey (i.e., the discovery phase), buying groups have already formed clear preferences and informal shortlists. Faced with this reality, marketing’s goal must be to influence these decisive buyers before they enter a formal purchase process, requiring committed, long-term investment to position the brand so that it is placed on the shortlist. This is a radical departure for many B2B companies that have built so-called “performance marketing” teams focusing almost exclusively on in-market buying groups that display intent. Marketers must rebalance resources from “performance marketing” and toward more proactive “preference marketing.” They must marry brand and demand marketing efforts and build sales and marketing playbooks based on preference positioning. Build Preference Marketing B2B marketing teams still need to find a balance between the near-term imperative of engaging with active buying group members and the longer-range goal of building preference in a market or segment so that it’s not one or the other. What’s clear is that marketing teams creating campaigns and programs that build preference will drive more sustainable long-term growth and higher levels of marketing effectiveness and ROI than those that focus on “performance marketing” alone. By making the building of brand preference a priority, you will help ensure that your organization is on the shortlist when buying groups begin their formal purchasing process, ideally as their preferred vendor of choice. Utilize market research, brand trackers, and other methods to monitor buyer sentiment and preference across your most relevant markets and product categories. This data, along with other preference metrics, will also show how building buyer preference drives a step-change improvement in downstream demand metrics, such as pipeline growth, win rates, and market share. The bottom line: Once buying groups enter their formal buying process and show a clear intent to purchase, they already have preconceived ideas about which vendors they favor and which they don’t. Your company will be too late to exercise significant influence on these decisive buyers and will likely end up as “cannon fodder” for pricing negotiations against the preferred vendors. That’s not a good place to be. Forrester clients can read our new research report on how to understand and win the preference race. Schedule a guidance session to adapt your strategy for how marketing and sales engage with decisive buyers. source

With clients struggling to nail their product launches recently, I thought it might be helpful to summarize a webinar I participated in a few months ago called Captain Your Launch: Mastering B2B Product Launches. Host Rowan Noronha was joined by Julien Sauvage, Tamara Grominsky, and myself. We covered the gamut on launches, sharing our experiences, examples, and best practices to help product/portfolio marketers create more impactful launches. Read on for the top takeaways from our discussion. Takeaway 1: Create overarching launch goals and plans, and break them down by different launch audiences. Just like any other initiative, launches should have a clear business objective (e.g., gaining share against a competitor, improving customer retention, expanding into a new market or category) with specific quantitative goals (e.g., X number of qualified opportunities by Q3). Consider the critical launch audiences, both internal and external, that are needed to make the launch successful. Internal audiences include sales, other customer-facing teams, and other colleagues — from executives to engineering. External audiences include customers, prospects, channel players, and influencers. Create launch plans with goal-oriented outcomes for each of these audiences that link to the broader launch goals and increase the probability of success. Takeaway 2: Segment your customers, and zero in on those with the strongest need. While companies tend to want to target the largest market segments, it is more important to find segments that have the strongest need for your offering and therefore have the greatest propensity to buy. Targeting any or all customers, or customers that are in a very large and broad market, can lead to situations where sales cycles take longer, might require price concessions, or result in churn down the road. As Rowan explained, you want to avoid “margin-sucking” customers who have a larger cost of acquisition, may not pay full price, may require extra servicing, and have short lifetimes. Use customer insights from primary and secondary research to identify the customer segments that have the strongest need for your offerings and that are willing to pay for them. This will help create higher average-per-customer revenue and increase lifetime value. Takeaway 3: Get feedback from potential customers throughout the launch process. Product managers and portfolio marketers should always speak to customers to get firsthand feedback on the product definition and go-to-market efforts for the launch. Trialing new concepts, MVPs, and beta testing earlier in the development process is necessary to understand the problem you are trying to solve, the product-market fit, and key features and capabilities. Conduct message and pricing testing to reveal the strongest value propositions, define packaging options, and crystalize the pricing strategy. For example, Tamara described using paid ads to test three different value propositions to identify which resonated the most with target customers; the results informed both the product roadmaps and their messaging, creating a strong response for the related launch. Takeaway 4: Structure launch teams based on the launch tier for faster time to market and higher impact. Launches can be complex projects involving multiple functions, people, and activities. To create the right level of impact without an excessively heavy process, product marketers should set up structured launch teams relative to the scope and tier of the launch. For bigger launches (tier 1 or 2), three teams might be required: 1) a cross-functional leadership group that provides executive sponsorship, makes strategic decisions, and handles escalation and resourcing issues; 2) a core team that executes the work (events, content, demand, communication, sales, etc.); and 3) an extended team that should be informed about important launch factors or decisions (e.g., finance, legal, compliance, operations). Lower-tier launches can often be handled by the product marketing and product managers and a few additional core team members and functional leaders as needed. Takeaway 5: Determine the most appropriate launch strategy. Launches can get their own go-to-market campaign or be part of a broader campaign with integrated efforts spanning reputational, demand, sales, and customer communication activities. You can use a classic big bang effort to unveil something new and exciting all at once, a rolling thunder approach with a coordinated series of activities over time, or limited releases to allow for rollout to different customer segments. Regardless of the launch strategy, Julien stated his inclination to use an external influencer — a thought leader, analyst, or customer — who provides third-party credibility and validates the claims you’re making for the launch. Takeaway 6: Create a one-page launch overview to get everyone “on the same page.” While detailed project plans ensure timely completion for the launch, Forrester has found that a one-page launch overview is extremely helpful for aligning everyone’s understanding of the launch. Lay out the following sections on one page to make it easy to read: the launch overview with tier, objective, and goal; the target audiences including target market segments and buyers; the offering overview with the value proposition, differentiator, packaging options, and pricing; and the launch strategy, including a timeline of activities and events leading up to the launch, at launch, and post-launch. Use this one-pager to align internal audiences, from executives and sales to product teams and engineers. Internal clarity helps create external clarity. Takeaway 7: Establish a launch retrospective to continuously learn and improve go-to-market efforts. No matter how much customer feedback you get in advance of a launch, bringing something new to market provides a wealth of new information. Are we finding our target buyers? Is the message resonating? Is the pricing on target? Are there new competitors we didn’t know about? Include a launch retrospective in your plan a few weeks or more after the launch to gather feedback from your launch team(s) on what worked, what didn’t, and what you should keep doing, stop doing, or change. Make sure to capture any unanswered questions and think about how you might get the answers. Launches continue to form a critical piece of go-to-market strategy for B2B organizations; successful launches can begin a trajectory for successful commercialization of new innovations and capabilities. Leverage these launch best

The Forrester Wave™: Security Analytics Platforms, Q2 2025, published today and illustrates the dramatic changes this market is undergoing as legacy security information and event management (SIEM) vendors are locked in heated competition with surging extended detection and response (XDR) providers. Over the past five months, we researched the top security analytics platforms on the market to gain a better understanding of the market landscape and identify the best product fit for our clients. This Forrester Wave evaluated 10 vendors: CrowdStrike, Elastic, Exabeam, Google, Microsoft, Palo Alto Networks, Rapid7, Securonix, Splunk, and Sumo Logic. Each vendor was evaluated on three inputs: a questionnaire for the vendors to complete, executive strategy briefings and demos, and interviews of three reference customers. The Wave included scores for 24 current-offering criteria and six strategy criteria. Read the full report here. Forrester defines security analytics platforms as: Platforms that converge data from network, identity, endpoint, application, and other security-relevant sources to generate high-fidelity behavioral alerts and facilitate rapid incident analysis, investigation, and response. This evaluation marks a turning point for the security analytics platform market. XDR vendors such as CrowdStrike and Palo Alto Networks have staked their claim for what they consider a new era of SIEM capabilities — one that is heavily focused on detection and response. The challenge with this vision is that, in contrast to many of the SIEM tools on the market, they tend to lack flexibility in data ingest, manipulation, and core compliance use cases. Some SIEMs have a longer head start, bigger communities, a wider variety of features, and an entrenched customer base. The biggest players of the industry — e.g., Google, Microsoft, and Splunk — have amassed an array of features that prioritize data, openness, and adaptability. With all of that said, customers have been frustrated with the SIEM market for a long time. Particular pain points include the quality of prebuilt analytics, the massive amounts of manual work required, and the high cost. This push and pull is leading to a shift in the market marked by the following trends: Flexibility versus specialization. Many of the veteran security analytics platforms have an understanding of data, from ingestion to manipulation and searchability. These platforms are indispensable for complex use cases that require flexibility; the trade-off is more work for the end user. Contrast that with the XDR vendors, which have limited collectors to focus on detection and response and more security-specific query languages, and they can also build out-of-the-box analytics over time. Both approaches have value — it just depends on what you want to get out of the tool. Platformization. Platformization. Platformization. Security analytics platforms (as the name suggests) are a natural fit for platformization. The security analytics platform is the central location where security operations takes place. This is why Forrester has defined it as the security analytics platform market (and not the SIEM market) since 2015. Despite efforts to deliver interoperable products with third parties, nothing integrates or bundles quite like native tools. To take advantage of the pendulum swing of platformization, some XDR vendors do not charge for the ingestion of their own endpoint detection and response (EDR) data, thereby saving practitioners money. Everyone’s favorite: Generative AI. Every vendor in this evaluation discussed its AI capabilities as part of a vision for the future. But the differentiation between vendors when it comes to AI was stark. Some vendors are pushing unique features out fast, while others are stagnating. Many vendors had some of the AI functionality we have come to expect: incident summaries, chatbots, and query language translation. The vendors that differentiated, however, delivered AI agents, automated parsing, and other leading features. AI will change the way security operations functions, and betting on the right horse now will enable your team to change with it. For a deeper look into the market, Forrester clients can read the full report, The Forrester Wave™: Security Analytics Platforms, Q2 2025. Check out the results for all 10 vendors, including the specific criteria that differentiated them and why. If you have any questions about the changes happening in the security analytics platform market, book an inquiry or guidance session with me. source

GUI, Chat, Voice, XR, Oh My! With the proliferation of social media platforms, smart devices, and conversational interfaces, the ways Europeans interact with brands have expanded. Customer touchpoints now span websites, mobile apps, chatbots, email, voice assistants, and even wearable tech — each presenting a unique opportunity to acquire, engage, and serve. Customers – more empowered, discerning, and willing to explore than ever – expect convenient, seamless, and value-driven experiences from brands. Meanwhile, CX and digital business leaders often voice frustrations such as “We launched this new feature, but hardly anyone is using it — and we don’t understand why” or “We implemented a chatbot, but engagement is low. How do we get customers to actually use it?” These challenges typically arise from a mismatch between the solution and customer’s actual needs. Sometimes, the tool fails to address the problem effectively — or worse, it’s solving the wrong problem entirely. Moreover, customer behavior is deeply influenced by context. People interact with digital tools differently depending on their environment, goals, and constraints in the moment. Digital Behaviors Continue To Evolve Customers are increasingly using voice commands through virtual assistants on smartphones or smart speakers to control devices and access information – valuing speed and hands-free convenience. Yet for tasks such as checking on a delayed delivery, those same users may prefer texting with a support agent via a chat interface. For more involved activities, such as consumption and commerce tasks, they often turn to a rich graphical user interface (GUI) on a smartphone — for example, to compare product reviews before making a purchase or transaction. These shifting preferences underscore how context — from environment to urgency — shapes digital behavior.   Understand The European Digital Consumer Landscape To meet rising customer expectations and design digital experiences that truly resonate, organizations must ground their digital strategies in a deep, data-driven understanding of their users. This starts with uncovering actionable insights into audience behaviors and preferences. To steer your digital experience (DX) strategy effectively, you need to understand which interaction modes (e.g., GUI, voice, chat, etc.) consumers use for various tasks (e.g., control, communication, commerce, etc.) and how many devices, channels, and platforms they rely on. That’s why we developed two powerful tools: the Digital Moments Map and the Digital Connections Tracker. Each combines a dataset and a model to help you make smarter DX decisions. The Digital Moments Map The Digital Moments Map assesses user behaviors and attitudes across 16 different combinations of interaction modes and task types based on an annual Forrester survey. It helps answer questions such as “What percentage of people in Germany prefer using voice for control tasks, such as adjusting a smart thermostat?” We’ve just published five European market-specific reports in this series to support these decisions: This task-based approach highlights the importance of designing experiences that are both context-aware — responsive to the user’s situation — and preference-driven, tailored to individual needs and expectations. The Digital Connections Tracker The Digital Connections Tracker assesses how many devices, channels, and platforms people rely on – by region/market — based on an annual Forrester survey. It helps answer questions such as “How many devices, channels, and platforms are people using on average this year in France?” These insights, for instance, can help you determine whether your marketing campaigns are reaching users across the right number of touchpoints. Here are the five European market-specific reports in this series: These reports reveal that European consumers are highly connected, engaging across an expanding array of devices, channels, and platforms — often simultaneously. A typical consumer might browse a product on a tablet, compare prices on a smartphone, and finalize the purchase on a desktop. This multidevice, multichannel, multiplatform behavior has redefined customer expectations; consistency and continuity across digital touchpoints are now a requirement. To meet these expectations, brands must deliver seamless, integrated digital experiences, regardless of where, when, or how users choose to connect. Develop A Data-Driven DX Strategy Together, these reports highlight a critical shift: DX strategies must evolve beyond product- and channel-centric approaches to become context-aware — understanding the nuances of digital moments — and channel- and device-agnostic. A deep understanding of where, how, and why people interact across digital channels enables organizations to: Adopt a customer-first mindset. Start with the end user and work backwards to inform strategy and execution. Prioritize with purpose. Use data-driven insights to focus on what truly matters, while confidently deciding what not to pursue. Design unified digital experiences. Create seamless, context-aware interactions that are connected, convenient, and aligned with user needs. By quantifying digital engagement patterns, organizations can confidently choose and justify their DX investments, enhance UX, boost engagement, and anticipate future needs. Let’s Connect Forrester provides clients with custom guidance and data. To get started, schedule a guidance session with me to discuss the models and data in more detail. You can also request custom data cuts segmented by country, year, age group, income band, and more — depending on your specific needs. source

When we published our report, The Age Of Agents — GenAI-Fueled Virtual Assistants Will Fundamentally Change Digital Interactions, in 2023, we were watching the rise of generative AI and looking around the corner, noting that AI agents will reduce consumer friction and change how consumers engage with digital businesses. Two years later, agentic AI is one of Forrester’s top emerging technologies. Meanwhile, consumers are going all in on various genAI applications (one of our family members has an ongoing thread with ChatGPT to help him rehab a sports injury). While agent-based scenarios are newer, consumers are starting to pay attention. In a recent Consumer Pulse Survey we conducted, 36% of US adults said that they would be somewhat or very interested in delegating an AI agent to find and book reservations for travel, concerts, and/or other experiences. That shot up to 53% when looking just at US adults who are part of Generation Z.   For consumer-facing businesses, this means that: Proven digital strategies will be challenged. Marketplace business models will struggle to aggregate more than a consumer agent. Price discrimination will be quickly identified and not tolerated. Comparison services will have a hard time remaining relevant when agents can review consumer-specific criteria and make specific recommendations in no time. CX will increasingly equal “agent experience.” Good CX is going to require good agent experience for an increasing share of online consumers. Consequently, the “psychology of CX” will change when agents enter the fold. Contrary to humans, they won’t be susceptible to habit-creating features, gamification, or gimmicky loyalty programs. Client relationships will shift. Agents are poised to become the main user interface for those who use them. Curation and personalization will be provided by the agent, not businesses, while most client data will remain with the agent. Domain-specific agents will be rejected. Several consumer-facing businesses are building their own proprietary “shopping agents.” But the power of consumer agents lies in access to holistic consumer data. A personal agent will always outperform a domain-specific corporate agent in gathering deep client insight, driving far superior personalization and curation. For Security And Fraud Teams, It’s No Longer As Simple As “Bot Or Not” Security, fraud, and e-commerce teams have adopted bot management tools to distinguish automated traffic from human traffic, block the bad bots, and let in the good bots and the humans, while also sometimes throttling good bots during business-critical events such as flash sales. This gets more complicated with agents, which might look like bots but will have a specific human being’s intent behind them. The line between human and automated has never been blurrier. Look for bot management offerings that quickly adapt to analyze agent traffic and determine intent to give businesses the information they need to prioritize that traffic appropriately. Digital And Security Teams Must Collaborate To Optimize Revenue And CX Cross-team collaboration will be more important than ever. Security, fraud, digital, and marketing teams must work together to determine how to manage and prioritize the combination of good agents, bad agents, good bots, bad bots, good humans, and bad humans now gracing their digital storefronts. This discussion will require nuance. If agents always manage to cut the line ahead of humans trying to access the site directly, then human consumers could get frustrated and go elsewhere, a scenario that has long been an issue in the bot world. Always blocking or slowing down agents, however, will frustrate the humans behind them, such as the 53% of US Generation Z consumers interested in delegating agents for travel, concerts, and experiences. Blocking agents will send those consumers elsewhere. Therefore, teams will have to work closely to develop agent interaction strategies that prioritize their business’s target consumers and improve CX for them. Later this year, we will be publishing additional research on the topic, including some more data points on consumer attitudes toward agents. In the meantime, we invite Forrester clients to set up an inquiry or guidance session to discuss this further. source

The US government is no stranger to cloud. In fact, the federal government was among the first adopters of public cloud going back to the first federal CIO, Vivek Kundra, who mandated a cloud-first approach to IT as part of the 2012 budget process. Although his motivation to adopt cloud was instrumental in transforming government IT, Kundra mistakenly assumed that it would save billions of dollars. The early promise and hyperscaler marketing hype of cloud as a cost-savings mechanism has been disproven. Despite this, the cloud is a transformation and innovation accelerator and a necessary part of modern IT strategies. The rush to the cloud, particularly following the pandemic years, led to an alarming spike in costs that required attention. Enter FinOps. The practice has existed for almost a decade but has only recently been adopted at scale — especially in the US government. Aligning Cultural Practice With Political Pressure Contrary to popular belief, government leaders have always been mindful of cost optimization. The Federal Acquisition Regulation is designed to prevent malfeasance, abuse, and waste, but there are gaps. In 2019, the US Government Accountability Office discovered that of the 16 agencies reviewed, roughly one-third had inconsistent reporting in cloud investments, even though most agencies had a 10-point or more increase in spending. These agencies had saved $291 million but had also identified issues in reporting and tracking cloud spend and savings. This was due, in part, to a lack of consistent processes. Today’s cost-cutting pressure and implementation of mass-scale efficiency, particularly with the Department of Government Efficiency (DOGE), is not new to IT leaders. FinOps practices and investment value maximization were already in play; DOGE just accelerated that mission. What has changed is the increased scrutiny of cloud spending decisions. This has traditionally been a safe haven for procurement, as public cloud contracts envelop not just the acquisition of cloud services but also the SaaS solutions that can be purchased through their marketplace. This has been displaced as DOGE increased scrutiny on SaaS spend. Still, FinOps practices are rising to this pressure. Up until recently, FinOps practice scopes did not encompass SaaS spend or any other costs outside public cloud infrastructure and services — this has changed. The new FinOps Foundation framework has changed in terminology to include cloud but also “technology.” Increased scrutiny on all tech spend will place a greater onus on FinOps teams to provide cost optimization outside of the public cloud. Overcoming The Structural Challenges Of Budgeting And Procurement Federal acquisition lifecycles are unlike traditional enterprise processes. Government entities typically drive procurement in the public space. Flexibility on spend and methods is highly limited, as these are generally based on earmarked money. In other words, public procurement budgets are more likely to be delegated preemptively and are therefore tougher to alter in distribution. In the federal space, the Antideficiency Act (ADA) presents a huge obstacle to cloud purchases. Agencies are required to obtain appropriations and declare their entire projected cloud consumption in advance. The ADA also prohibits obligating or spending money before it’s received from Congress and does not allow the redistributing of funds to a different purpose than the one declared. This limitation provides little resilience, if any, on managing spend anomalies. It also means that overcommitment is a common occurrence. Securing Savings Through Commitment And Consolidation To combat this, government FinOps teams will employ tactics such as commitment-based discounts (e.g., reserved instances, reservations, savings plans, and commitment-use discounts). Cloud cost management tools are also in play, though common burn-down tactics of penalties on overages are not allowed with these contracts, thus limiting options on tools. Some agencies struggle to benefit from the cost levers available to commercial-side teams. While the federal government is a big spender in IT, its efforts are typically federated, and individual PMOs rarely have the negotiating power to move the needle for big tech negotiations. Without the prerequisite purchase volume or procurement flexibility, substantial committed use discounts are a pipe dream. One proven model that works (for agencies with the executive buy-in to pull this off) is a consolidated procurement. The DoD’s Joint Warfighting Cloud Capability vehicle is a leading example of government cloud. By aggregating requirements across multiple branches of the military, the DoD was able to negotiate discounts and favorable terms that would not likely be available to individual PMO buyers. And like other verticals, government teams should also implement the key tenets of FinOps: individual accountability (though showback is more common, since chargeback is easily resisted), cross-functional collaboration (though standard procedures must be followed versus organic watercooler conversations), timely and accurate decision-making, and cost. For more detail on implementation and execution, Forrester clients can use the Forrester Solution Blueprint, Optimize Your Cloud Costs With FinOps. source

Huawei just raised the stakes For those just tuning in, NVIDIA has been driving much of the AI infrastructure conversation. Huawei has been, not so loudly, building its own AI stack, right from the silicon to systems to AI models. CloudMatrix ups Huawei’s play in the market considerably. The CloudMatrix-Infer system is no ordinary cluster. It brings together 384 Ascend 910C NPUs and 192 Kunpeng CPUs, interconnected via a Unified Bus (UB) with ultra-high bandwidth and low latency that challenges NVIDIA NVLink. CloudMatrix impressed us with its: System-level integration optimized for inferencing at scale. Huawei is proving its vertical stack (Ascend + CANN + MindSpore + UB) with tight coupling/integrations. It supports disaggregated prefill-decode-caching architecture, INT8 quantization, microbatch pipelining, and fused operators, boosting performance for inferencing. Chipset advantage of Ascend 910C and Kunpeng. Although the single-chip performance for Ascend 910C is about 33% lower than NVIDIA’s Blackwell, by networking five times more chips via a 2.8Tbps UB, Huawei has turned a weakness into dominance. On the other hand, each node pairs Kunpeng CPUs with Ascend, handling control-plane tasks like distributed resource scheduling and fault recovery. This hybrid design ensures non-AI workloads like data preprocessing and network management don’t bottleneck NPU efficiency. Unified Bus vs. NVLink: topology as a weapon. While NVLink focuses on ultra-fast point-to-point GPU connections, Huawei’s UB implements an all-to-all interconnect topology by using 6,912 optical modules to weave a “flat” network of all 384 NPUs and 192 CPUs across 16 racks. This eliminates hierarchical hops and enables direct communication. CANN support of massive-scale expert parallelism (EP320). This allows one expert per NPU die —something even CUDA ecosystems struggle with. While NVIDIA’s CUDA dominates global AI development, Huawei’s Compute Architecture for Neural Networks (CANN) stack has quietly matured into a viable alternative. Now at version 7.0, CANN mirrors CUDA’s layered structure across driver, runtime, and libraries, optimized for Ascend. Throughput over raw speed for performance advantage. CloudMatrix has 6,688 tokens per second prefill and 1,943 decode per second per NPU — better than NVIDIA H100 on similar loads. Developer-oriented updates. Recent improvements to the PyTorch compatibility layer, as noted in the paper, suggest that Huawei is listening to its early developer base. Huawei is optimizing for massive mixture of experts (MoE) models, bandwidth-first design, and multiphase LLM inference at scale. If I were deploying a 700B+ model in a region where CloudMatrix was available, we’d seriously consider using CANN. The architecture is purpose-built for next-gen inference. Watch Huawei closely, even if you still lean on NVIDIA Huawei isn’t just catching up with its CANN stack and the new CloudMatrix architecture, it’s redefining how AI infrastructure works. I believe winning the AI race isn’t just about faster chips. It also includes delivering the tools developers need to build and deploy large-scale models. As someone who’s built applications (in my past life), I’d rely on CUDA’s mature, frictionless ecosystem. I can run PyTorch and TensorFlow code with minimal tweaks, tap into global community support, and use robust tooling that just works. Huawei’s CANN software development kit shows potential — but it still feels early. PyTorch and TensorFlow adapters exist, but migration takes effort, and the community remains small. It’s not turnkey yet, and that slows developer confidence. Huawei has its work cut out Despite the gains, Huawei has lots of work to do to improve: Power efficiency. CloudMatrix consumes 3.9 times more power than NVIDIA’s GB200. This may be viable in China’s abundant, less expensive utilities market, but it’s a big hurdle in regions where energy is expensive or carbon targets are strict. Optimization is a must. Ecosystem maturity. While CANN is making progress, CUDA still leads in documentation, third-party libraries, and global developer community. Migration friction. Tools and adapters exist — but real-world code migration still takes effort. This is a pain point for teams that need fast iteration cycles. Community trust. Open-source engagement and hands-on support must scale. Especially outside Asia, trust and familiarity favor NVIDIA. Some companies and public sector groups may not be legally able to use Huawei given its relationship with the Chinese government. For the Chinese public sector, however, Huawei is already the de facto choice. We see these items as solvable, but they’ll take time and focus. Final thought NVIDIA’s market share in China is facing challenges. With ongoing geopolitical uncertainty, diversifying infrastructure is no longer optional — it’s a strategic imperative. Tech leaders should accelerate localization strategies, and infrastructure is critical for AI success. To dive deeper into your AI strategy, set up an inquiry or guidance session with Charlie Dai (AI cloud) or Naveen Chhabra (AI infrastructure) for a conversation. source

In the ever-evolving circus of enterprise tech and AI advancements, Datadog’s 2025 keynote was clearly fueled by the AI arms race. CEO Olivier Pomel and his team unveiled a dizzying array of AI-powered tools and capabilities promising to transform the mundane world of IT operations from a reactive firefighting exercise into a proactive, almost magical realm of autonomous problem prevention and remediation. From AI security analysts that triage threats in seconds to dev agents that generate pull requests while engineers sleep, Datadog is betting big on the promise that AI will rescue us from the soul-crushing minutiae of modern software development. The overwhelming list of announcements that lasted for nearly 2 hours ended with a screen filled with nearly 100 items that were spoken to during the keynote. Whether these collectively materialize into a revolutionary leap forward or not remains to be seen, but one thing is certain: The future of tech looks increasingly algorithmic. The Promise And Peril Of AI-Driven Observability Datadog’s DASH 2025 keynote was a masterclass in AI optimism — equal parts compelling and questionable, some elements now generally available but many still at “preview” or “coming soon” status. Bits, Datadog’s AI, featured prominently in a multitude of product announcements and enhancements. Bits gains an AI voice interface and will have broader data access and exploration capabilities to improve root-cause analysis. As an example, Bits AI SRE was positioned as an autonomous incident resolution AI agent that can simultaneously investigate multiple root causes in minutes. The APM Investigator promised to deliver intelligent pattern recognition based on memories and learning. The Dev Agent will work with error tracking, Real User Monitoring, traces, databases, test optimization, and more. Datadog’s AIOps and observability platform gets enhancements grounded in the expansion of capabilities to extend its reach both deeper into technologies and broader into the organization for contextual awareness. While all these announced capabilities are intriguing, there will be many cultural, in addition to technological, challenges to overcome in the real-world complexity of IT and technology platforms before becoming mainstream. The narrative conveyed was that of the Datadog platform being the central hub for operational intelligence, a notion promoted by more and more vendors in this market space in 2025. Missing from DASH, as if nonexistent, was any conversation about AI hallucinations, the computational cost of these “intelligent” systems, and the human expertise that these tools might inadvertently deprecate. Datadog Continues Its Push Into Security Operations Datadog announced security monitoring capabilities in 2019, marking its first entry into the security information and event management (SIEM) market. Since then, it has grown its capabilities and team significantly — one of its more recent announcements in December 2024 highlighted its approach to risk-based insights and its threat detection and engineering team. This year at DASH, it unveiled its Bits AI Security Analyst, a parallel agent to the Bits AI SRE and dev agents it also released. Much like other recent AI agent announcements from competitors such as Microsoft, CrowdStrike, Google Cloud, and others, the AI Security Analyst automatically investigates SIEM alerts within its domain expertise. For Bits AI Security Analyst, that currently includes AWS CloudTrail, with expansion into other domains coming soon. Datadog also announced enhancements to log management that affect its SIEM and observability offerings. It introduced a Flex Frozen storage tier to retain logs up to seven years, and, importantly, the ability to archive search without re-indexing, which is helpful for compliance use cases. Other vendors like Elastic have similar features, such as searchable snapshots to be able to access archived data quickly. Datadog also announced a CloudPrem capability to deploy Datadog index and search on-premises to meet data residency and compliance mandates. Forrester continues to see a push and pull from clients that want the freedom of a cloud deployment versus the potential regulatory requirements that necessitate an on-premises deployment, and this feature will hopefully help strike the balance between the two with what was previously a capability solely available in the cloud. Datadog Releases Features To Help Teams Secure AI Datadog also made a plethora of announcements on securing AI deployments, especially with regards to securing the models themselves. It announced Datadog LLM Observability, which is generally available and allows users to trace large language model (LLM) chains and troubleshoot outputs, among other features such as performance monitoring and metrics measurement. It also announced Datadog Workload Protection, which received an upgrade that’s currently in preview, that handles LLM Isolation, enforcing guardrails on deployed LLMs and identifying vulnerabilities. And lastly, Datadog Code Security is now generally available for the identification and prioritization of vulnerabilities in custom and open-source code, integrated into the developer’s integrated development environment. When it comes to artificial intelligence security, Forrester recommends focusing first on the users and prompts before moving into model security itself — many of the attacks on models are, at this point, academic, but there are many vendors exploring LLM security at this point. SIEM vendors are in some ways uniquely positioned to support emerging use cases such as LLM security, as data can be ingested from these tools quickly and users can build specific analytics to meet their use case. We recommend scheduling a guidance session with Jeff Pollard for more in-depth conversations on which vendors may be the best fit. Data Intelligence Underpins Operational Insights Datadog’s data intelligence approach is its response to the growing complexity of enterprise data ecosystems, seeking to address this challenge through data observability. Clients should watch this area closely for end-to-end data lineage tracking as the capability matures over time. You will want to see how these solutions provide visibility across the entire data lifecycle — from source systems through transformation to consumption — a critical pain point for data engineers and analysts. The ability to trace data issues upstream and understand their broader impact is a desired capability by enterprises, but it is a challenging one fraught with obstacles for vendors. AI agent interaction monitoring represents an emerging need for organizations that are increasingly

Cisco Live 2025 Focused On Three Main Themes: AI, Simplification, And Security At its annual Cisco Live event, the company delivered a clear message: It’s operationalizing AI across the core pillars of networking, security, and observability. Building on last year’s momentum with innovations like Hypershield and Splunk integration, the company has framed its vision around helping organizations become AI-ready by embedding intelligence into infrastructure, simplifying operations, and securing increasingly complex environments. Forrester analysts observed that Cisco is focused on efforts to do the following. Accelerate AI Adoption And Transform IT While this continues the narrative that Cisco began last year, the vendor has leaned heavily into its core DNA: its dominant network infrastructure presence. This year, however, the emphasis shifted to helping organizations make AI work for them. Cisco is now doubling down on making AI an operational advantage rather than an existential risk through a wave of targeted announcements, including: Accelerating troubleshooting with AI. This arguably became the star of the show. In a nutshell, AI Canvas is the purpose-built, AI-based collaborative workspace that allows for advanced troubleshooting within an innovative UI. The idea here is for both NetOps and SecOps teams to work together with AI agents to diagnose, troubleshoot, and (ideally) proactively resolve complex network security issues. It will also be a core feature within Cisco’s latest Security Cloud Control (the product formerly known as Defense Orchestrator) and underpins Cisco’s “AgenticOps” narrative. IT value: accelerates troubleshooting and reduces downtime through intuitive, AI-driven diagnostics Security value: enables proactive resolution of complex network security issues with AI agents Building AI-ready data centers. Jeetu Patel, president and chief product officer, highlighted the need for enterprises and service providers to build out their current infrastructure to support their unique AI workloads. To support this transformation, Cisco announced the following: AI PODs built on NVIDIA GPUs combined with Cisco UCS; Cisco G200 switches and NVIDIA NICs; NVIDIA- and Cisco-validated designs; and new 400G optics. IT value: scalable, high-performance infrastructure tailored for AI Security value: secure, future-proof hardware for on-premises and hybrid data centers Injecting AI into the workplace. From Webex to Cisco’s network management platforms, AI assistants and agents will help organizations ensure that infrastructure is always on, transport more traffic, and support Zero Trust. AI will help organizations shift from reactive to proactive approaches across campus, branch, and industrial networks. IT value: ensures always-on infrastructure and proactive network optimization Security value: enhances Zero Trust and visibility across campus, branch, and industrial networks Secure The AI Infrastructure Tom Gillis, senior VP of infrastructure and security, emphasized during an analyst roundtable that clients should “use AI to protect AI.” Cisco backed that up via several announcements that aim to deliver innovative hardware and software solutions to future-proof data centers and equip them for AI adoption. Here is a quick breakdown of those announcements. Secure Routers and Smart Switches. The Cisco Secure Routers (8100, 8200, 8300, 8400, and 8500) will now combine native SD-WAN integration, L7 firewall capabilities, and post-quantum security within a single appliance. The vendor’s smart switches include its latest Nexus 9300 and Catalyst 9350 and 9610, which will come equipped with dedicated data processing units courtesy of its AMD partnership. IT value: reduces hardware sprawl and simplifies operations. This allows the switches to seemingly take the place of a rack-mounted firewall, reducing costs associated with maintaining more equipment without jeopardizing the security of your data center. Key to this is Hypershield, which promises to deliver AI-driven security architecture across a distributed network for security enforcement. Security value: delivers firewall-grade protection at the switch level, without added complexity or cost Enhanced Secure Firewall portfolio. Cisco continues to improve upon its Secure Firewall, focusing on its integration with the broader Security Cloud suite, coupled with AI-driven capabilities. It announced two new models: the 200 series for branch offices and the 6100 series for data centers. Along with Hypershield, its Secure Firewall (both hardware and software instances) will contribute to Cisco’s “Hybrid Mesh Firewall” push, which also includes Meraki and (oddly enough) ASA. IT value: streamlined deployment across hybrid environments Security value: supports Cisco’s Hybrid Mesh Firewall strategy with AI-driven enforcement across Meraki, ASA, and more Future-proofing the data center. Cisco’s announcements aim to secure data centers regardless of the hosting environment, especially as more organizations repatriate workloads on-premises. IT value: scalable, future-ready infrastructure that supports hybrid workloads and simplifies data center modernization. Security value: high-performance, security-centric hardware that protects assets without inflating capex. As more organizations repatriate some of their assets back on-premises, sectors outside of telcos and service providers now require high-performance, security-centric hardware to secure their evolving data center requirements. Build Platforms To Simplify Operations Cisco has one of the largest networking and security product portfolios. Historically, this portfolio was fragmented, difficult to implement, and complex to manage. Over the last 10 years, the company has fallen behind the market in enterprise networking, Wi-Fi, security appliances, and UC&C, along with its recognition as an innovator and thought leader declining. As Forrester highlighted in its Cisco Is Getting Back Into The Game blog, Cisco wanted to change its image and announced that it was merging the Catalyst and Meraki product lines. Simplifying operations. Cisco’s Security Cloud Control aims to consolidate network and security management into a single platform, driven by its native AI Canvas and AI assistant. This seeks to reduce complexity and operational overhead by allowing for consistent policy management across Cisco solutions, leveraging a “mesh” policy engine that will even apply to third-party firewalls. The vendor is also promising that there will be no need to rip and replace current investments. IT value: reduces operational overhead and training requirements Security value: enables consistent policy enforcement across campus and branch networks Security Cloud Control and AI-powered automation. The vendor aims to have its platform leverage AI to optimize policies, detect and identify network issues, and even recommend solutions. It also will seek to enhance the ability to automate routine tasks and support improved threat detection and response. It will do