Preface Executive Briefing

Week of June 6, 2026

Tokens are the receipt. Not the result.

As AI adoption grows, the question is shifting from how much are we using to what work is actually getting done.

The gap

The bill is visible. The value is not.

When organisations start tracking AI, they usually begin with the easy things to count: licences, token spend, cloud cost, and tool usage. The harder question is what completed work or business value those signals actually point to.

The visible bill

Licences, tokens, cloud spend, GPU capacity, vendor subscriptions.

The hidden value

Avoided labour, faster decisions, less rework, lower marginal cost.

The reframe

Tokens prove the machine ran. Not that work got done.

Most AI dashboards blur three separate things: what went in, what happened, and what work was actually completed.

  1. Input

    Tokens, API calls, compute, licences

  2. Activity

    Prompts, agents launched, documents generated

  3. Outcome

    Work completed, risk cut, time saved, value captured

Most dashboards stop after the first two rows.

The trap

Measure usage, manufacture fake transformation.

When usage becomes the target, people maximise usage, not value. That is AI theatre: visible activity, no redesigned work.

What you measure

Usage is climbing

Tokens consumed

+340%

AI sessions per week

+210%

Leaderboard rank

+4 positions

Documents generated

+180%

Everything looks great.

What you do not ask

Outcomes are missing

Customer problem solved?

Cycle time reduced?

Quality improved?

Cost per outcome?

Nothing answered.

Reported examples at Amazon and Microsoft, and more tentatively Meta, point to the same risk: visible AI activity can look like progress even when the underlying work has not changed.

The shift

Service-as-a-Software: the recommended solution.

Once AI can complete parts of a workflow, the old software question becomes less useful. The better question is not who needs a seat, but what work should be completed, to what standard, and at what cost.

Tool model

Buy the software

The company pays for seats, licences, or usage. People still operate the product and the value is inferred from adoption, activity, or productivity gains.

Easy to track

Who used the tool?

Outcome model

Buy the result

The customer pays for a completed workflow. AI handles the repeatable steps, people stay on exceptions and judgement, and success is tied to whether the work was finished well.

Harder, but better

What work changed?

This is the strategic move behind Service-as-a-Software: AI makes it possible to sell the finished work, not just the interface people use to get there.

The proof

Define the unit before you claim the value.

A token count does not tell you whether the work improved. A completed unit does. That is how hidden AI value becomes something finance and operations can inspect.

Per resolved ticket

Customer support handled end to end, with verification and escalation only when needed.

Per reviewed contract

Standard contracts reviewed against approved playbooks, with deviations flagged and escalated.

Per reconciled invoice

Invoices ingested, classified, reconciled, and prepared for approval.

Per qualified lead

Leads researched, scored, and prepared with verified contact and intent data.

These units are the bridge from activity to value: they turn AI from something the organisation consumes into work the organisation can verify.

Key Takeaways

The question is no longer how much AI are we using

It is what work did AI complete, and did it create measurable business value?

01

Tokens prove activity, not value

Tokens show that a machine processed something. They do not prove that valuable work was completed. The executive job is to distinguish spend, activity, and completed work.

02

Usage metrics can manufacture fake transformation

When usage becomes the target, people optimise for visible activity instead of better outcomes. Use adoption metrics diagnostically, but judge AI by what work changed.

03

The next value model sells outcomes, not access

Service-as-a-Software becomes credible when the unit of value is explicit: a resolved ticket, a reviewed contract, a reconciled invoice, a qualified lead.

FAQ

No. Most AI vendors still price by tokens, seats, or API calls. Outcome-based pricing is the emerging direction, not the established norm. The briefing explains why the shift makes strategic sense, not that it has already happened.