Walmart’s fiscal 2026 bets: Supply chain optimization, AI, automation

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Walmart continues to bet on leveraging AI, optimizing its supply chain and developing a host of related businesses that'll grow operating profits.

The retail giant reported fourth quarter earnings of 65 cents a share on revenue of $180.6 billion, up 4.1%. For Walmart, it's the first time its quarterly sales were below Amazon sales. For fiscal 2025, Walmart reported earnings of $2.41 a share on revenue of $681 billion.

Walmart projected fiscal 2026 net sales growth of 3% to 4% with operating income growing from 3.5% to 5.5%. Capital expenditure will be between 3% and 3.5% of sales as Walmart invests in technology to optimize its supply chain, remodel stores and open new ones.

Here's a look at what Walmart CEO Doug McMillon and CFO John David Rainey had to say on the earnings conference call.

Walmart has introduced Wally, a new AI agent for the company's merchants. "Wally is learning to help us get to the root cause of issues related to things like out of stocks or overstocks with more accuracy and speed," said McMillon.

Developers at Walmart are leveraging AI for coding and deployments and the company saved 4 million developers hours last year. This year, all of the AI coding tools will be available to all developers in North America and India.

• Sam's Club CEO Nicholas on AI, frictionless commerce, focus on members
• Walmart's big tech, AI bets paying off going into 2024 holiday shopping

Walmart's complementary businesses will grow operating profits. The company's global advertising business delivered annual sales of $4.4 billion, up 27% from a year ago. Walmart US Marketplace revenue was up 37% and 45% of orders were filled by Walmart Fulfillment Services. And global membership income was up 21% to $3.8 billion for fiscal 2025. These businesses will grow operating profits faster than sales.

Walmart data is becoming a business. Rainey said: "Within data analytics and insights, Walmart Data Ventures continues to grow rapidly with net sales up double digits. Our client base nearly doubled over the past year, and we're excited about continuing to broaden our reach to new markets with the launch of the platform in Canada."

Walmart's business is 18% e-commerce today and the percentage will increase in the future. Since the selling, general and administrative expenses (SG&A) related to an e-commerce transaction are higher than physical retail, Walmart is betting on automation. Rainey said:

"As you think about our cost structure going forward, one of the big drivers is going to be the improvements that we see in supply chain automation. We're already seeing that. We're encouraged by some of the early productivity metrics. But still today, less than half of the stores in the U.S. are served fully by automation. And so, there's a lot of benefit still to come here as we automate our supply chain as we continue to automate our stores that will drive improvements in SG&A."

Walmart will highlight its investments in supply chain automation in April at its investor conference, said McMillon.

The retailer is honing its delivery game with same-day pharmacy delivery and various shipping offers at Walmart and Sam's Club. McMillon said Walmart is taking the lessons from markets such as China and applying them elsewhere.

Walmart is also trying to lower its cost to serve. Rainey said drivers are aiming to deliver to more houses on a street, add volume and create more dense networks.

Walmart's return on investment improved 50 basis points to 15.5%, a level that was last achieved in 2016.

The company is prepped for a potential volatile economy and customers are looking for value. Rainey said: "Our outlook assumes a relatively stable macroeconomic environment but acknowledges that there are still uncertainties related to consumer behavior and global economic and geopolitical conditions. As a result, we've taken a similar approach to our initial guidance view for the year as we have in the past couple of years, balancing known risk with what we can control. We remain confident that Walmart is well positioned to navigate as it has over the last several years."

Tariffs. Like most enterprises, part of the potential supply chain volatility revolves around tariffs. McMillon said: "tariffs are something we've managed for many years, we'll just continue to manage that. We've got a great team. We know how to do that. We can't predict what will happen in the future, but we can manage it really well. And we're wired to try and save people money. So that will be our ultimate goal."

Alibaba's cloud unit garners Q3 AI demand boost, touts Qwen efforts

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Alibaba's cloud computing business delivered 13% revenue growth and the company said it is "committed to advancing multi-modal AI technology and expanding our opensource initiatives."

While DeepSeek models have garnered the attention, Alibaba has continued to advance its Qwen large language models (LLMs). For the enterprise, Qwen may be more scalable and a better option than something with a bunch of unknowns like DeepSeek.

DeepSeek's real legacy: Shifting the AI conversation to returns, value, edge

Alibaba's Cloud Intelligence Group reported fiscal third quarter revenue of $4.35 billion, up 13% from a year ago. Excluding Alibaba subsidiaries, Alibaba's cloud unit delivered growth of 11% from a year ago. Adjusted EBITA for the third quarter was $430 million.

The surge in demand "mainly driven by the double-digit revenue growth of public cloud products including AI-related products," said Alibaba. The company also noted that AI-related product revenue has maintained triple-digit revenue growth for six consecutive quarters and will continue to invest in AI infrastructure.

Alibaba noted:

“We remain committed to advancing multi-modal AI technology and expanding our open source initiatives. In January 2025, we open-sourced Qwen2.5-VL, our next-generation multi-modal model, and launched our flagship MoE-based model Qwen2.5-Max. Both models deliver globally leading results across recognized benchmarks and are available to users and enterprises through Qwen Chat and our Bailian platform. Since August 2023, we have open-sourced various large models under the Qwen family. As of January 31, 2025, more than 90,000 derivative models had been developed on Hugging Face based on the Qwen family of models, making it one of the largest AI model families worldwide."

Fiverr's grand AI experiment: Melding creatives with personalized models

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

For Fiverr, artificial intelligence isn't about replacing creative labor. The plan is to meld digital and human labor in a way that's not zero sum.

Fiverr, a marketplace for freelancers, launched Fiverr Go, an AI platform that enables creators to tune models based on their portfolio of work. The idea behind Fiverr Go is to enable creators to use AI to refine workflows and customer experience, and deliver mockups and first drafts based on their style.

The Fiverr take on AI is notable because it is less zero-sum than what's being pitched by most of the tech industry. Let's face it. Technology giants are increasingly talking about digital labor as replacing humans. And if you follow the money--actually it's more like layoffs--human worker concerns about AI are justified.

This AI digital labor game puts a company like Fiverr in a pickle given generative AI can replace freelancers--and the need for its platform. Fiverr CEO Micha Kaufman positioned Fiverr Go this way on the company's fourth quarter earnings conference call:

"Different from other AI platforms that often exploit human creativity without proper attribution or compensation, Fiverr Go is uniquely designed to reshape this power dynamic by giving creators full control over their creative process and rights. It allows freelancers to become a one-person production house, making more money while focusing on the things that matter creating."

With Fiverr Go, freelancers can create personalized AI models without collecting datasets or understanding the engineering. Fiverr's dataset of 6.5 billion interactions and 150 million transactions on its marketplace round out the training.

Kaufman said AI results are generic and hard to edit and many customers are using Fiverr freelancers to refine images and results. "Every delivery on Fiverr Go is backed by the full faith of the creator behind it, with an included revision as the freelancer defines. This means that the quality and service you get from Fiverr Go is no different from a direct order from the freelancers themselves," said Kaufman.

Fiverr Go's Personalized AI Assistant can communicate with potential clients and handle routine tasks. The Fiverr Go AI Creation Model can create mockups and first drafts based on the freelancer's work. Kaufman said the company also launched a Freelancer Equity Program that will give shares of Fiverr to top-performing freelancers.

The bet is that Fiverr Go, which launched with 60 categories, can convert sales faster for freelancers. The personal AI assistant "is essentially encapsulating the entire knowledge of the freelancer and basing itself on it being able to address any possible question and bring it to conversion," said Kaufman.

In addition, the AI creation model "allows customers to get the confidence that this is the freelancer, this is the style that they're looking for" without the back-and-forth of delivering samples and essentially working for free, said Kaufman.

The Fiverr Go experiment is one worth watching.

Fiverr's outlook

Fiverr said it is expecting to grow revenue at double-digit rates in 2025 with sales of $422 million to $438 million, up 8% to 12%. For the first quarter, Fiverr projected revenue of $103.5 million to $108.5 million.

Kaufman said Fiverr has services revenue momentum going into the first quarter.

For the fourth quarter, Fiverr reported net income of $12.8 million, or 33 cents a share, on revenue of $103.7 million, up 13.3% from a year ago. Marketplace revenue was down 4% from a year ago with 3.6 million active buyers at the end of 2024, down from 4 million a year ago.

For 2024, Fiverr reported net income of $18.2 million, or 48 cents a share, on revenue of $391.5 million, up 8.3% from 2023.

More on digital labor:

• Anthropic's Economic Index highlights AI augmenting mid-to-high salary jobs
• Workday aims to be system of record for AI agents, digital labor
• Agentic AI without process optimization, orchestration will flop
• CxOs upbeat on economy, plan to invest heavily on genAI, AI agents
• Agentic AI: Three themes to watch for 2025

Box launches new AI features, Box AI Units

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Box launched new AI features for Enterprise Advanced customers and said it will add consumption-based pricing via Box AI Units.

The moves highlight how SaaS vendors are increasingly moving toward hybrid seat-subscription-consumption models with generative and agentic AI. This consumption model also lands as large language models (LLMs) are commoditizing at a rapid clip.

• AI agents bring consumption models to SaaS: Goldilocks or headache?
• DeepSeek: What CxOs and enterprises need to know
• DeepSeek's real legacy: Shifting the AI conversation to returns, value, edge
• GenAI prices to tank: Here’s why

Under Box's consumption model, the company will introduce Box AI Units, a metric that tracks and manages AI API usage. The transparency can be used to scale and optimized without unpredictable costs. Also: Box acquires Alphamoon Technology, aims to integrate LLMs, OCR into Box AI

In addition, Box said its customers on Business, Business Plus and Enterprise plans will have unlimited AI querying for docs and images. Customers on Individual, Personal Pro and Starter plans will have Box AI capabilities in the months ahead.

As for the features, Box launched the following:

• Box AI Extract Agents, which automate metadata creation and provide insights from documents and images. Extract agents are built into Box experiences natively.
• Custom Agents, which can be created via Box AI Studio API. Box AI Studio launched last month.
• Multi-Doc Querying, which queries multiple documents at once to help with research, legal and business analysis.
• Box Forms and Box Doc Gen. Box Forms gives customers the ability to collect information and start business processes via web and mobile forms. Once the data is collected, workflows can be automated for onboarding, service requests and claims processing. Box Doc Gen gives customers the ability to generate unlimited custom documents natively in Box, Salesforce and via API.

Box's AI strategy revolves around embedding AI throughout its platform and multiple layers including metadata, user experiences and workflows.

Microsoft unveils Majorana 1, aims to scale quantum computing

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Microsoft launched Majorana 1, a quantum computing chip with a Topological Core architecture.

Topological quantum computing uses a concept similar to semiconductors using "anyons," which can arrange qubits into patterns. A topological superconductor is a material that can create a new state of matter. It's harnessed to create a more stable qubit that can be digitally controlled.

According to Microsoft, Majorana 1 has a breakthrough material that can observe and control Majorana particles to create more reliable and scalable qubits. Chetan Nayak, Microsoft Technical Fellow, said the goal for Majorana 1 was to invent "the transistor for the quantum age."

Microsoft is betting that Majorana 1 will be a more fault tolerant way to scale quantum computing. Microsoft's architecture used in Majorana 1 creates a path to fit 1 million qubits on a chip the size of a palm of a hand.

• Quantum Computing Platforms
• Quantum Computing Software Platforms
• Quantum Full Stack Players

There are various flavors of quantum computing in addition to the approach Microsoft is using:

• Superconducting qubits are seen as general quantum computing options and vendors in this category include IBM, Google and Rigetti Computing.
• Trapped Ion quantum computing has high fidelity and long coherence times. IonQ is the big player in this category along with Quantinuum, which was created by the merger of Honeywell's quantum unit and Cambridge Quantum.
• Neutral atom quantum computing has the potential to scale better and QuEra is a player here.
• Quantum annealing is designed for optimization over general purpose computing and D-Wave has championed this approach.

Microsoft said Majorana 1 has eight topological qubits on a chip and can scale from there. Microsoft is building its own hardware as well as partnering with the likes of Quantinuum and Atom Computing.

• Quantinuum, Microsoft claim quantum reliability breakthrough
• Microsoft claims hybrid quantum breakthrough with Quantinuum, partners with Atom Computing

Years or decades?

Quantum computing has been in the middle of a big debate about whether it'll be useful in years or decades. Nvidia CEO Jensen Huang said in January that quantum computing was 15 to 30 years away from being useful. Microsoft said its approach will scale quantum computing "within years, not decades."

• Taking stock of quantum computing’s kerfuffle and what's next
• CxOs need to focus on quantum computing readiness, not the noise
• Practical quantum computing advances ramp up going into 2025
• Quantum computing all in on hybrid HPC with classical computing

Microsoft outlined Majorana 1 in a paper in Nature.

Under a program with DARPA, Microsoft said it will build the world's first fault-tolerant prototype based on topological qubits.

Nayak said Microsoft's plan now revolves around "making more complex devices" including its first QPU Majorana with a topological core. Nayak said Microsoft "can scale to a million qubits on a chip the size of a watch face."

Given Microsoft's developments, the move by Quantinuum to combine quantum and generative AI and various hybrid HPC and quantum efforts, enterprises need to prepare potential use cases.

• Quantinuum launches generative AI quantum framework, sees quantum computing as synthetic data generator

Indeed, cloud vendors, which will deliver quantum instances, have been busy setting up services to get enterprises quantum ready. Given the mileposts, quantum computing is developing quickly.

Constellation Research analyst Holger Mueller said:

"Right when you think quantum computing approaches were set we have a new approach and vocabulary to learn with topological qubits and majorana. It is good to see that alternate approaches are feasible, promising and could accelerate the path to quantum, but it does leave a few question marks for other quantum vendors scaling out alternate approaches. What will matter for CxOs will be a consistent software layer across platforms to traverse quantum platforms. But first we need to see the viable quantum platforms."

 

Airbnb: With tech stack in place, expansion plans accelerate

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Airbnb plans to become a platform for travel akin to how Amazon in commerce courtesy of a multi-year transformation and a new tech stack.

Speaking on the company's fourth quarter earnings call, Airbnb CEO Brian Chesky said the company has spent the "past several years preparing for Airbnb's next chapter" and rolled out more than 535 features in upgrades in its app over the last two years.

Indeed, that velocity is due to a revamped tech stack designed to take advantage of AI and improve the overall experience. Airbnb is largely built on Amazon Web Services and has nearly 5 billion visitors a year.

Chesky said Airbnb has launched Guest Favorites, which makes it easier for guest to find listing, the Co-Host network, to find local hosts to manage your Airbnb, and destination and map improvements. For good measure, Airbnb is redesigning its checkout experience to remove friction.

These improvements have boosted conversion rates to drive fourth quarter results. Airbnb reported fourth quarter net income of $461 million on revenue of $2.5 billion, up 12% from a year ago. For 2024, Airbnb reported net income of $2.6 billion on revenue of $11.1 billion, up 12% from a year ago.

Chesky in February 2023 noted that AI would benefit Airbnb's long-tail of data. Since Airbnb's 2020 IPO revenue has tripled. In many ways, Airbnb appears to be a travel version of Uber, which has a model that revolves around data.

• Uber outlines its autonomous vehicle plan
• How Uber's tech stack, datasets drive AI, experience, growth
• Uber aims to press its 'data advantage' for AI, model training

Chesky said:

"We rebuilt our platform from the ground up with a new technology stack. We've also upgraded our messaging system into a single unified platform, making communication between guests and hosts smoother more reliable. Now, with this new tech platform, we are able to innovate faster and expand beyond short term rentals into becoming an extensible platform with a range of new offerings and 2025 marks the start of Airbnb's next chapter."

Simply put, Airbnb plans to expand beyond its core business to adjacent areas starting in May. Airbnb plans to invest $200 million to $250 million to launching and scaling new businesses. A lot of that spending will go to product development and marketing.

The AI strategy

With the new tech stack, Chesky said Airbnb wants to be in position for trends like agentic AI, but noted that "it's still really early."

"I think AI is going to have a profound impact on travel, but I don't think it's yet fundamentally changed for any of the large travel platforms," said Chesky. "We want to be the leading company for AI enabled traveling and eventually living."

Chesky noted that trip planning with AI is early and not ready for prime time. "We're actually choosing a totally different approach, which is we're actually starting with customer service. So later this year, we're going to be rolling out as part of our summer release, AI powered customer support," said Chesky, who noted genAI works well with multiple languages and can read thousands of documents easily.

Airbnb CEO: GenAI's impact on app experiences minimal so far

Going forward, the plan for Airbnb is to take that AI customer service agent and bring it to Airbnb search to "eventually graduate to be a travel and living concierge," said Chesky.

Chesky said AI will drive efficiencies for customer service as well as developer productivity.

With model prices falling, the commoditization of models will drive value to platforms. Chesky noted that Airbnb aims to be the traveling and living platform.

In the big picture, Chesky said that Airbnb can leverage AI effectively because it is one brand and one app.

He said:

"We want the Airbnb app, kind of similar to Amazon, to be one place you go for all of your traveling and living needs. A place to stay is just really, frankly, a very small part of the overall equation. Every new business we launch, we'd like to be strong enough. It could stand alone, but it makes the core business stronger. I think that each business could take three to five years to scale. A great business could get to a billion dollars of revenue. Doesn't mean all of them will. And you should be able to expect, like, one or a couple businesses to launch every single year for the next five years. We're going to start initially with things very closely adjacent to travel."

Chesky added that new businesses for now will stay close to travel and then Airbnb will expand from there. For now there are dozens of adjacent markets for Airbnb to expand into.

The tech stack

Although Airbnb has been a relatively vocal customer of AWS in recent years it has been quiet. Airbnb was a featured customer at AWS re:Invent 2022, but has noted in its annual report that it is busy integrating AI into its tech stack.

Airbnb noted in its annual report that "our technology platform incorporates the use of AI and ML Technologies, for example, for fraud detection, search, enabling customized features and enhancing community support."

Chesky seemed to indicate that Airbnb was focused on multiple models and optimization, which would point to services like Amazon Bedrock. DeepSeek and cheaper models will also bring down prices, he noted. "I think it's a really exciting time in the space because you've seen like with DeepSeek and more competition with models is models are getting cheaper or nearly free, they're getting faster and they're getting more intelligent and for all this purpose, starting to get commoditized," he said.

Airbnb said in its annual report that it is investing in developing, maintaining and operating AI and machine learning models. That investment will also increase compute costs in the future. Airbnb is also investing in developing its proprietary data sets.

Beyond the AWS foundation, Airbnb is decidedly has a build culture focused on open source tools, sponsoring them and building frameworks that can be used to continuously improve and integrate new technologies.

Notable upgrades include the following:

• An Intelligent Automation Platform (AP) designed for conversational AI. AP is an enterprise platform designed to support conversational AI products across multiple channels.

• Airbnb's Data Warehouse storage from legacy HDFS clusters to S3 and a conversion to Iceberg tables.
• Graph neural networks.
• And Journey Platform, a low-code tool to create complex workflows.

AI agents bring consumption models to SaaS: Goldilocks or headache?

Larry Dignan

Editor in Chief of Constellation Insights
Constellation Research

Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads editorial coverage focused on enterprise technology, digital transformation, and emerging trends shaping the future of business. He oversees research-driven news, analysis, interviews, and event coverage designed to help technology buyers and vendors navigate complex markets with clarity and context. ...

Read more

Enterprise procurement departments are already annoyed with software-as-a-service contracts and AI agents--and the consumption-based models that go with them--are likely to make deals even more complicated.

Welcome to the new world of enterprise software--licenses, seats and a heavy dose of credits and consumption charges. Get ready for conversations like the following:

CFO: "Our IT operating expenses are running hot."

Procurement: "Yeah, we were dinged by extra AI packs, power ups and consumption units."

IT: "We need to optimize our AI agent costs. These $2 conversations with AI are adding up."

Within a few weeks you can rest assured that your enterprise software providers will be layering in consumption models. On the bright side, CIOs are used to consumption models from their hyperscale cloud providers including AWS, Microsoft Azure and Google Cloud as well as data platforms such as Snowflake and Databricks. The bad news: CIOs have struggled to manage those cloud consumption costs for years.

In recent weeks, we've seen the following:

• Monday.com will introduce AI Blocks, Product Power-ups and Digital Workforce. These tools are designed to streamline and automate processes with AI. Monday.com will offer a generous free tier and then you can buy AI credits from there.
• ServiceNow outlined its AI agent pricing with a hybrid approach that blends seats, subscriptions and consumption. CEO Bill McDermott said the model is a "Goldilocks scenario" for ServiceNow and customers. "Customers can start with a base subscription, which they like. They want that flag in the ground, so they can predict their spend and their current ROI schemes. But then, they obviously want to take advantage of agentic AI and yet at the same time, the industry is early in its formation," said McDermott.
• SAP outlined AI and data strategy as well as a partnership with Databricks that will likely include consumption pricing.

HubSpot CEO Yamini Rangan explained on the company's fourth quarter earnings call that the company's approach has been to create an AI-first product without add-ons for AI. Monetization has come by raising prices for the overall product. Going forward, hybrid models based on seats and usage will be the norm for AI agents.

"I do think that the future of pricing for AI will be hybrid. We'll have both seat based and usage based pricing. Right now, we're focused on delivering value with our agents and as more customers get consistent value with AI, we will introduce usage based pricing," explained Rangan. "The pricing model will be a combination of usage and seats based pricing. But what is really important is that we will consistently focus on delivering value first before adding on to our seat-based models and then monetizing based on usage."

And those are just some recent examples. Salesforce monetizes Agentforce via consumption pricing, Adobe has sold credits for its AI usage and Dynatrace, Confluent and others are going the same route. You can expect a consumption announcement from a SaaS vendor almost weekly going forward.

What does this mean for the enterprise buyer?

Consumption will surge. These consumption models from software vendors often include a healthy free tier because they want more usage. Once enterprises move past free tiers there will be a learning curve to optimize costs and AI agent use cases. Remember how shocked companies were when they thought the cloud cut costs? Get ready for the SaaS-y version. The good news is that DeepSeek and cheaper models will also bring down prices.

• How leaders need to think about AI, genAI | Enterprise software 2025: Three big shifts to watch

Transparency will be at a premium. AI agents are going to require pricing transparency that SaaS vendors aren't used to providing. Enterprise software vendors will need to provide the same dashboards and consumption dashboards that hyperscale cloud players deliver. One customer, who is an early adopter of AI agents, said he expects that same cost transparency from his SaaS vendor as he gets from AWS. If anything, SaaS vendors should have better transparency.

• Agentic AI: Three themes to watch for 2025 | Agentic AI without process optimization, orchestration will flop

Cloud marketplaces will be critical. Efforts like AWS Marketplace that enable enterprises to purchase software and roll up procurement under one dashboard will become popular. Procurement is already coming around to cloud marketplaces and consumption transparency will accelerate that move.

• Big software deals closing on AWS Marketplace, rival efforts | AWS Marketplace adds 'Buy with AWS' as it expands reach, woos procurement departments

It's unclear who will manage the digital labor force. One news item that was notable this week was the Workday Agent System of Record. The big idea is that Workday already manages human capital and it can extend into digital capital, aka AI agents too, and track returns and onboarding.

Attribution of outcomes will be the missing link in AI agent consumption models. Ron Miller, operating partner and head of editorial at boltstart ventures, said on DisrupTV that figuring out what vendor is responsible for an outcome is going to be messy. Miller said: "If you start talking about outcome pricing as another element of consumption-based pricing it's chaotic. Who was responsible for the outcome? Was it the Salesforce piece? Was it the Box piece? Was it the ServiceNow? That's just another piece of all this."

Customers will demand cross-platform AI agent transparency. Miller's take revolves around the harsh reality of AI agents today: Every vendor thinks enterprises only operate on one platform. The reality is that there will be AI agent workflows that may be connected from AWS to Google Cloud to Salesforce to Workday to ServiceNow. How do you optimize that mess when each vendor has different pricing? AI agents will make those per-core pricing schemes look straightforward.

The Goldilocks scenario will be delayed. McDermott obviously thinks that the seat, subscription and consumption hybrid model is a win-win for vendors and customers, but I'll bet that there will be a lot of grumbling on the way to Goldilocks.