On 24 July the US Department of Veterans Affairs awarded Salesforce a contract with a ceiling of 1.6 billion dollars. The headline number travelled fast. The structure underneath it deserves more attention, because the deal was signed as an Agentic Enterprise License Agreement, and Salesforce AELA changes the maths on how agent programmes get planned, funded and defended internally.
The agreement runs as one year with two further one-year options, and it covers agent capability for a population of more than 17 million veterans. The Salesforce-powered contact estate underneath it has already handled north of 40 million calls since it went live. This is not a pilot that grew up. It is a decade-old CRM estate being extended into agents under a single commercial envelope, which is a far more common starting position than the greenfield builds that dominate conference keynotes.
That envelope is the genuinely interesting part, and the news is better than most of the commentary suggests. For the past two years the practical brake on agent adoption was rarely model quality or platform readiness. It was arithmetic. That arithmetic just got considerably easier to do, and the organisations that notice first will spend 2027 building while everyone else is still modelling.
What Salesforce sold the VA was a licensing model, not just software
Salesforce Agentforce is Salesforce’s platform for building and running autonomous AI agents inside the Customer 360, where an agent can reason over CRM records, call defined actions, and complete a task such as triaging a case, verifying a benefit, or surfacing the right answer to a caseworker during a live call. Until recently it was sold largely on consumption, with Flex Credits drawn down per conversation or per action. The Agentic Enterprise License Agreement, or AELA, wraps Agentforce together with Data 360, MuleSoft and Slack into a single multi-year enterprise agreement priced around committed user volumes rather than per-interaction metering. The VA award is the largest public example of that structure so far.
It is not the only one. The Adecco Group signed a multi-year AELA in March covering operations in more than 60 countries. Two very different organisations, one a federal health and benefits agency and the other a global staffing firm, arrived at the same commercial shape within months of each other. When buyers that unalike converge on a structure, the structure is usually solving a real problem rather than a marketing one.
Predictable pricing removes the wrong kind of caution
When every agent conversation carries a visible marginal cost, teams ration. They pick the highest-volume, lowest-risk use case, they cap the rollout at a department, and they quietly never find out what agents are good at across the long tail of smaller tasks. That caution feels prudent in a steering committee. It is also expensive, because the long tail is precisely where most manual effort actually sits: the fifty low-volume processes that each consume a few hours a week and none of which would ever survive a business case on their own.
Flat commercial envelopes change that calculation. The question stops being what you can afford to automate and becomes what you have the process clarity to automate. That is a much better question, because it is one your own team can answer without a procurement cycle, and because the answer improves every time you document a process properly.
A caveat worth stating plainly: seat-based does not mean unmetered. Most vendors, Salesforce included, pair per-user licensing with underlying fair-use thresholds, credit allocations or premium tiers that keep model costs in check at the extremes. The meter has not been switched off. It has been moved out of the daily build queue and into the contract, which is a different thing and, for planning purposes, a considerably more useful one.
Why Salesforce implementations go over budget, and what actually changes here
Salesforce implementations go over budget for reasons that are rarely about the licence. The recurring causes are scope discovered late, because processes were never documented before configuration began; customisation built to preserve a legacy way of working rather than to improve it; data migration that turns out to be a data quality project wearing a disguise; and integration work sized from an architecture diagram rather than from the actual state of the source systems. Licensing is the most visible line item, so it absorbs most of the scrutiny, while the real variance sits in discovery, data and change management.
A flat agent licence removes one genuine source of budget variance, and that is worth having. It does nothing whatsoever to the other four. Any organisation treating a predictable agent price as a substitute for process documentation will find the overrun simply relocates. The opportunity is to spend the certainty well: use the fixed envelope to fund the discovery and data work that consumption pricing always made hard to justify, because nobody wants to pay per conversation to find out whether a process is worth automating.
How to measure CRM ROI when the licence is already paid for
Measure CRM ROI by comparing the fully loaded cost of platform, implementation, integration and ongoing administration against a small number of operational measures the CRM can plausibly move: cycle time from qualified lead to closed order, win rate on contested deals, cost to serve per case, forecast accuracy, and the hours per week that sellers and service staff spend on administration rather than on customers. Take the baseline before you change anything, because retrospective baselines are always flattering. Under a flat agent licence the cost side becomes fixed for the term, so the honest question shifts from cost per interaction to how much work you genuinely moved onto the platform while you were paying for it.
That shift rewards breadth. Under consumption pricing, the rational move was to automate narrowly and defend the credit burn. Under a committed envelope, an agent handling a low-value process at low volume costs nothing extra and still returns time to the people doing that process today. Utilisation becomes the metric that decides whether the agreement was a good one, and utilisation is entirely within your control.
Adoption becomes the variable that decides the value
Improve CRM adoption by making the system the fastest route to doing the job, rather than an additional place to record that the job was done. In practice that means removing fields nobody reads, designing screens around the actual sales or service motion, giving managers reports they genuinely run their weekly meeting from, and appointing named process owners who can approve changes in days rather than quarters. Train on the workflow, not on the software. Measure adoption through behaviour, such as the share of open pipeline updated within the last fortnight, rather than through login counts, which tell you almost nothing.
Agents make this easier in a way that deserves more credit than it gets. A great deal of low CRM adoption has always been a rational response to administrative burden: people avoid the system because the system asks them for things that do not help them sell or serve. When an agent drafts the summary, updates the record after the call and chases the missing field itself, adoption improves for the plainest possible reason. There is less to adopt.
What to ask before you sign a multi-year agentic agreement
An independent CRM consultant is paid to recommend the right platform and the right scope, including the option of buying less. A vendor’s own professional services team is genuinely excellent at deploying that vendor’s product, but its commercial incentive points towards more product, more modules and larger commitments. On a multi-year agentic agreement that difference matters most at the moment you size the commitment, because the honest number is frequently smaller than the modelled one. An independent partner also works across Salesforce, HubSpot and Dynamics 365, so any comparison of what each platform actually does well rests on delivery experience rather than on a single vendor’s catalogue.
Three questions are worth asking before signature. Ask what happens to consumption transparency once per-interaction billing disappears, because you still need to know what your agents are doing even when nobody is invoicing you for it. Ask for post-term price protection in writing, since advisory firms consistently report annual uplifts in the high single digits to low teens and initial discounts that quietly evaporate at renewal. And ask to structure the deal as a conservative base with pre-agreed expansion options rather than a large day-one commitment, because first-year utilisation against committed pools routinely lands well below what the original model predicted.
Salesforce has said publicly that it is willing to lose money on early agent licences. That is worth reading exactly as it is meant. The value is priced across the life of the relationship rather than the first invoice. This is not a trap, it is a negotiating position, and negotiating positions work in both directions for a buyer who has done the arithmetic and knows what utilisation they can realistically reach.
The Sirocco perspective
We think the shift to committed agentic agreements is, on balance, good news for buyers, and we would rather say so than perform scepticism. Predictable pricing is the single change most likely to move agent work out of perpetual pilot and into production, because it removes the argument that stalls most steering committees. The organisations that benefit will be the ones that treat the fixed envelope as permission to be ambitious about scope rather than as a reason to stop thinking about it.
Where we would counsel care is in the sizing conversation, and only there. The commitment you sign should reflect the adoption you can credibly deliver in the first eighteen months, not the adoption a model projects for year three. That is a question about your processes, your data and your change capacity, and it is answerable in a few weeks of honest work. Our experience across Salesforce, HubSpot and Dynamics 365 is that clients who do that work first sign smaller, use more of what they signed for, and negotiate the renewal from a much stronger position.
If you are weighing an AELA or any multi-year agent commitment and want an independent view on the number before you sign it, schedule a consultation and we will walk through the sizing with you.
Get in Touch
If you are sizing a multi-year Agentforce or AELA commitment and want an independent read on the number before it goes to signature, tell us where you have got to.
