Home » Learning Center » Converting AI capability into measurable business value without paying twice 

19 minutes

Converting AI capability into measurable business value without paying twice 

In April 2026, enterprise AI capability moved from option to inheritance. The question for the boardroom is no longer whether to buy AI. It is whether the AI investment already made will become a measurable business outcome or sit on the shelf. This guide sets out the delivery model that closes that gap. 

Download

Table of contents

Contact person

Stef Knaepkens

Chief AI Officer

The cart-without-a-meal problem

Enterprise AI spending is up. Enterprise AI results are not.

In April 2026, the industry analyst Melody Brue of Moor Insights & Strategy described the state of enterprise AI in a phrase that landed in many boardrooms: artificial intelligence today is like a cart full of groceries without a meal to make. The metaphor stuck because it described what business buyers were already seeing. Licenses had been bought. Pilots had been run. Implementation partners had been engaged. Yet the link from capability to outcome was missing.

The MIT NANDA research initiative, the Media Lab’s program studying enterprise AI adoption, put a number on the problem. In its State of AI in Business 2025 report, NANDA found that 95% of enterprise generative AI investments have returned zero measurable business value. The conclusion was blunt: the barrier to scaling AI is not infrastructure, not talent, and not regulation. It is learning. Most organizations cannot close the loop between an AI behavior, the human response to it, and the next iteration fast enough to build trust.

ServiceNow’s April 2026 commercial change made the capability question moot. Under the new packaging, AI is no longer an add-on. It is built into every tier of every product—Foundation, Advanced, and Prime. Foundation supports employees with AI-assisted help. Advanced automates entire workflows end to end. Prime fields autonomous AI specialists capable of doing role-based work without supervision.

The procurement question is therefore answered. The strategic question is not. For most enterprises, the missing piece is no longer the technology. It is the delivery model, the way that owned capability is converted into operating result.

We call it the Activity Tax: the predictable, billable, and largely invisible cost of paying for motion rather than outcome.

The Activity Tax shows up in familiar ways:

A six-month implementation that ends with a steering committee asking what the business is now doing differently and getting no clean answer.
A roadmap of “AI use cases” with no baseline measurement of what they were meant to improve.
A supplier whose financial model improves as the scope expands, irrespective of whether the customer’s metrics move.
A renewal conversation in which the supplier defends the work, the customer defends the spend, and no one in the room can point to the result.

The Activity Tax is not the fault of any one supplier. It is a structural feature of how professional services have been bought and sold for two decades: in days, in sprints, in resources, in hours. The unit of trade is effort. The risk of the outcome sits entirely with the customer.

That model worked when systems were predictable and the deliverable was a system. It does not work when the deliverable is a business outcome and the system itself is probabilistic.

The four shifts

Moving past the Activity Tax is not a procurement adjustment. It is a change at four levels: rhythm, process, system, and mindset.

Rhythm: from linear deployment to clockspeed 

Traditional enterprise software is deterministic. Specify it, build it, test it, ship it, and it behaves the same way every time. AI does not. AI is probabilistic by design. The same prompt can return different answers. The same workflow can hesitate, misroute, or escalate depending on context the system has not seen before. Treating AI like traditional software — build it, ship it, declare done — produces exactly the result MIT NANDA documented: a high adoption rate and a near-zero value rate.

Plat4mation’s delivery model is grounded in the MIT NANDA principle that AI value is unlocked by short, fast feedback loops between AI behavior and human psychology. We call this rhythm clock speed. Every release of AI behavior into the workplace is observed against two reactions — what the AI did, and how the user, the supervisor, and the customer responded to it. The next iteration is tuned to both. Trust is not announced. It is earned, week after week, by small, visible adjustments that show the AI is learning the organization’s context and the humans are learning the AI’s limits.

Linear deployment optimizes for go-live. Clock speed optimizes for the moment a user starts to trust the system.

Clock speed is what makes the difference between a Day 90 outcome that holds, and a Day 90 outcome that decays the moment the consultants leave the room. The rhythm of the engagement is therefore the first shift, not an afterthought.

Process: from sprints to settlements 

The unit of delivery changes from a fortnight of work to a quarter of demonstrated improvement. Sprints still exist inside the engagement, but they are not what the customer is buying. The customer is buying the result those sprints are meant to produce. A baseline is captured at the start. A short, time-bounded engagement — typically ten to twelve weeks — runs at clock speed. At Day 90, the outcome is measured against the baseline, and the financial settlement follows the measurement.

System: from time-and-materials to performance bands 

The commercial structure changes. Instead of a rate card, the contract carries a performance band: a fixed price for hitting the agreed outcome, a defined consequence for falling short, and a defined upside for exceeding the target. The band is set before the work starts, against independent industry benchmarks the supplier and customer agree on. There is no surprise at the end. There is no debate about whether the work was done. The evidence is the result.

Mindset: from delivery to outcome

The hardest shift is cultural. For the customer, it means accepting that the supplier is no longer a vendor of capacity; the supplier is a co-owner of the outcome. For the supplier, it means accepting financial exposure to a result they used to negotiate around with the words “scope change.” For both, it means treating the engagement as a joint commitment rather than a transaction.

This fourth shift is the one that most often fails. Procurement teams can rewrite a contract in a week. Mindsets take a quarter to land—and a year to stick. Any supplier promising to make this change without addressing the cultural piece is selling a contract, not a model.

Pay for the meal, not the groceries

Plat4mation’s proposition is direct: pay for the meal, not the groceries.

Translated into a working delivery model, this becomes what we call the Outcomes Compact—a three-part agreement between supplier and customer that replaces the timesheet as the unit of trade.

Component 1 — the Baseline 

Before any sprint runs, two or three measurable starting points are captured. The discipline that matters is balance: one or two of those goals should be efficiency outcomes — time, cost, volume handled per person — and at least one should be a customer experience or quality outcome — first-contact resolution, employee satisfaction with a service, defect rate, time-to-onboard. The balance prevents the engagement from collapsing into a cost-cutting exercise, and it ensures that the AI investment is judged the way the board judges every other investment: against value created as well as cost removed.

Whatever the metric, it is measured, agreed upon in writing, and signed by both sides. No baseline, no engagement.

Component 2 — the Band 

A target outcome is set against each baseline. Around it sits a performance band: the price of delivery at target, the financial consequence of falling short, and the upside for exceeding. The band is bound by independent industry benchmarks, so the conversation never devolves into a debate about ambition. With two or three baselines, the band is structured as a portfolio—underperformance on one outcome can be partially offset by overperformance on another, which keeps both sides honest and committed across the full set.

Component 3 — the Review (and continuous monitoring)

At Day 90, the actual result is measured against each baseline using evidence the platform itself produces. Not a slide. Not an estimate. The audit trail. Settlement of the performance band follows that measurement.

The Day 90 review is the visible checkpoint. What sits underneath it is more important: continuous monitoring of three things—AI behavior, business outcome, and regulatory compliance running throughout the engagement, not waiting for the quarter to close.

Behavior monitoring tracks what the AI is doing in the workplace: where it is acting confidently, where it is escalating, where it is being overruled by users, where its accuracy is drifting. This is the clock speed signal—the loop that tells the team what to tune next.

Outcome monitoring tracks the business metrics against the baseline in real time. If an outcome is trending below band by Week 6, the team has six weeks to correct course rather than discovering the gap at settlement.

EU AI Act compliance monitoring tracks classification, documentation, human oversight, and incident reporting against the obligations attached to the system’s risk category. For European customers, this is no longer optional, it is operational.

Continuous monitoring turns the Day 90 review from an audit into a confirmation. Both sides arrive at it already knowing the answer.

The Compact answers the question every Chief Financial Officer asks of every consulting partner and rarely gets answered cleanly:

What does success look like, who is on the hook for it, and how will we know—every week, not just at the end?

In an Outcomes Compact, success is defined in advance in two or three balanced metrics. The hook is shared. The evidence is built into the platform and monitored continuously. The conversation moves out of the meeting room and onto the dashboard.

Why this works now

Four conditions have come together in 2026 that make outcomes-based delivery viable for AI work where, until now, it was confined mostly to infrastructure and run-rate services.

The capability is already paid for 

With AI now embedded into every ServiceNow tier, the customer is no longer being asked to fund a separate AI project on top of their platform spend. The capability is on the balance sheet. The delivery question is how to operationalize it. This is precisely the question Outcomes-Based Delivery is built to answer.

The measurement is automatic 

Modern platforms produce their own evidence. Every action is logged. Every workflow is timed. Every resolution is traceable. The data needed to settle a performance band — and to feed the clock speed loop — is no longer extracted by survey or estimate. It is generated by the system itself, and it is auditable.

The regulatory environment now demands continuous oversight 

The EU AI Act, in force since August 2024 and phasing into full effect through 2026 and 2027, changes the compliance posture for any AI system operating in a European context. High-risk systems require documented oversight, risk management, post-market monitoring, and incident reporting. A linear deployment model — build, ship, walk away — is no longer compatible with the obligations attached to the systems being deployed. Continuous monitoring is not a Plat4mation preference; it is increasingly the legal baseline.

The maturity is there 

After two years of pilots, proofs-of-concept, and quiet write-offs, executive sponsors are no longer looking for AI demonstrations. They are looking for AI results. They are willing to consider commercial models that put the result, not the activity, at the center of the contract.

The convergence of these four conditions is what makes 2026 the right year to retire the Activity Tax.

The proof

A delivery model is only as credible as the work it has produced.

Plat4mation has delivered more than 50 AI engagements under variants of the Outcomes-Based Delivery model. The program with Philips is among the references most often discussed in our customer conversations—a global enterprise operating across multiple regions, where the standard of evidence is high and the appetite for unmeasured spend is low.

The pattern that emerges across our engagements is consistent:

Outcomes are realized inside a quarter, not a fiscal year. When the engagement is structured around a measured Day 90 settlement and run at clock speed, the team finds a way to put a working outcome in front of the business within six to eight weeks.

The two-or-three-goal discipline holds. Engagements that mix at least one efficiency outcome with at least one customer-experience or quality outcome consistently produce a stronger board narrative than single-metric engagements. The balanced set is harder to game and easier to defend.

Scope discipline improves on both sides. When the supplier’s revenue depends on a measurable result, scope creep becomes a shared risk rather than a customer problem. Discussions about what to include and exclude become honest.

The second engagement is easier to commission than the first. Once an executive sponsor has seen a Day 90 settlement based on platform evidence—and a continuous-monitoring record they can hand to auditors and regulators — the appetite for continuing under a managed, longer-term arrangement is materially higher. The first outcome is the proof. The second is the relationship.

A note on what this model is not. Outcomes-Based Delivery does not eliminate cost. It does not remove the need for skilled people. It does not turn delivery into a button. What it does is move the conversation about value from the end of the engagement to the beginning — and put the supplier on the same side of the table as the customer for the duration.

Plat4mation has run this model across more than 50 AI engagements. The first outcome is the proof, and the second is the relationship.

What this means for you and how you buy services

If the model in this paper is the answer, the procurement function is where the answer first meets resistance — for a simple reason. The way most enterprises buy professional services today was built for a different kind of work.

The headline metric is the hourly rate. Three vendors are asked the same question. Three rate cards come back. The lowest one is held up as evidence of a good deal. The engagement begins. The decision was reached in days. The result, if it comes, will be argued about for months.

Hourly rate is a precise answer to the wrong question.

It measures how cheaply a supplier is willing to put a body in a room. It tells you nothing about whether the meter on your business metric will move. A vendor with a low hourly rate can bill a smaller invoice for delivering no outcome. A vendor with a higher hourly rate, working under a performance band against a measured baseline, can be the cheapest way to move the metric you care about—because the spend is tied to the result, not to the seat.

Hourly rate measures how cheaply someone is willing to be in the room. It does not measure whether the meter on your business metric will move.

Deliverable or outcome?

A deliverable is something the supplier controls. An outcome is something the business experiences.

A fixed-price deliverable project says: we will build you a configured workflow, a set of automations, and a training program. When those things exist and are accepted, we are done and you pay. The supplier controls whether they are built. Success is a signature on a completion document.  An outcome says: the average time to resolve an employee request will fall from four hours to ninety minutes. When that is true at Day 90, we settle. The supplier controls nothing about that number except the quality of their thinking and their delivery. The customer’s adoption, culture, and management attention feed into it too. Risk is genuinely shared.

Fixed-price deliverable models are not wrong. They are right when what is being bought is a construction — a system built to a specification. The problem arises when the construction model is applied to AI. AI behavior is probabilistic. It needs tuning, iteration, and trust-building with real users over time. A specification-and-acceptance contract treats the probabilistic as if it were deterministic. That is the structural failure in most projects that don’t make it to production today.

A deliverable answers the question: was it built? An outcome answers the question: did the business change? In 2026, boards are asking the second question.

Five questions worth asking a vendor

For AI work in 2026, the questions worth putting to a prospective supplier are different from the ones procurement teams have asked for the last twenty years. They are:

1. Will you commit to a measured outcome — and what is the baseline you will measure against?

2. What is the performance band — the price at target, the consequence for falling short, the upside for exceeding?

3. How will you monitor AI behavior, business outcomes, and EU AI Act compliance through the engagement — not just at the end?

4. What is your cadence — are you running at clock speed, with weekly observation of how the AI is behaving and how users are responding?

5. What is your track record of settling a performance band on a comparable AI engagement? Show me the evidence, not the slide.

If a vendor cannot answer those five questions on the first call, the hourly rate is the only answer they have. That is now useful information.

Are you ready to buy this way?

Ready does not mean every team is aligned and every internal process is rewired. Ready means some specific things are in place—most of which are inexpensive to establish if there is executive backing.

An outcome sponsor.  A named executive who owns the business metric you are trying to move, not just the project. This person carries the result to the board. If no one in the room is willing to do that, the engagement will drift.

Access to evidence.  Outcomes-Based Delivery is settled on data the platform produces. That data has to be accessible to the joint team. Restricting it on grounds of historical caution undoes the model.

A baseline-first commitment. Resist the urge to “just start”. The engagement begins when the baselines are measured and signed — not before. Two or three balanced goals, at least one on efficiency, at least one on customer experience or quality.

Procurement at the table early. Performance bands, continuous monitoring obligations, and EU AI Act compliance language need to be in the contract from the first draft. Procurement involved at the end becomes procurement undoing the model at the end.

Internal participation. This is not a hands-off arrangement. Your supervisors and users are part of the clock speed loop. Their honest feedback on AI behavior is what makes the model work. Build that expectation into how the engagement is announced internally.

Preparing for joint success

A short readiness conversation held before any commercial work starts which surfaces most of the gaps. We ask sponsors three questions:

Who, by name, will sign for the outcomes on Day 90?
What does your organization believe today about the value of AI — and what evidence would change that belief?
Are you willing to be wrong, in writing, about a baseline — and let the platform evidence settle the disagreement?

The answers to those three questions tell us — and tell you — whether the relationship is ready to operate on a different commercial footing, or whether a smaller pilot of the delivery model itself is the safer first step.

The point is not that every customer should switch tomorrow. The point is that the hourly rate, used as the decisive procurement metric for AI work, has reached the end of its useful life. Boards will ask, in 2026 and onward, what was bought and what moved. Hourly rate cannot answer either question.

What this means for your board

With the procurement reset in place, the conversation can move from how services are bought to how value is reported. If you are responsible to a board or executive committee for the AI capability your organization now owns, three questions are likely already sitting on your desk.

Where is the value? 

AI is now embedded in platforms your organization pays for. The next board meeting will ask what business metric has moved as a result. Outcomes-Based Delivery produces an answer with two or three numbers, baselines, and an audit trail — covering both efficiency gained and customer experience improved.

Who carries the risk? 

Under traditional engagement models, the customer carries the outcome risk and, increasingly, the regulatory risk. Under Outcomes-Based Delivery, the supplier carries a meaningful share of the outcome risk through the performance band and shares the regulatory monitoring discipline through continuous oversight. When you are accountable to a board, that explicitness — financial and regulatory — is worth a great deal.

How fast can we move? 

A ten-to-twelve-week engagement run at clock speed, with a measured outcome at Day 90, fits inside a single quarter. Most boards can absorb a quarterly cadence of measured outcomes. Few will continue funding annual programs with deferred answers.

Our recommendation to C-level sponsors is straightforward. Identify two or three outcomes you would like to move within a quarter — at least one efficiency, at least one customer experience or quality. Ask your supplier to commit to them under a performance band, with baselines measured before work begins,  behavior and compliance monitored continuously, and the result audited at Day 90. Make those outcomes the unit of trade.

If your supplier cannot, or will not, work that way, you have learned something important.

Next Step

Enterprise AI has reached the end of the era in which capability could substitute for outcome. The capability is bought. The capability is on. The question is whether your delivery model will let you collect on it—and whether your supplier will move at the clock speed AI requires to earn trust, hold compliance, and deliver value inside a quarter.

Plat4mation built the Outcomes-Based Delivery model for this moment. We carry a share of the result. We measure against the evidence the platform itself produces. We monitor behavior, outcomes, and EU AI Act compliance continuously. We settle at Day 90. And we have done it more than 50 times.

The cart is full. Let’s make a meal.

Plat4mation’s outcome leads work with C-level sponsors to identify the two or three board-ready outcomes that should be the unit of trade for the next quarter — and the baselines, performance band, and monitoring frame that go with them. This conversation typically takes 45 minutes.

To open this conversation, contact your Plat4mation account lead or visit plat4mation.com.

Sources cited:
MIT NANDA, State of AI in Business 2025 (the GenAI Divide), MIT Media Lab, July 2025; Melody Brue, Moor Insights & Strategy, April 2026; ServiceNow commercial model announcement, 9 April 2026; Regulation (EU) 2024/1689 (the EU AI Act), in force from 1 August 2024.

Join 1400+ ServiceNow professionals!

Sign up to our monthly Flow@Work Exclusive newsletter to get free access to our expertise and lots of tips and tricks to make work flow on the Now® Platform.