Why This Matters Now

AI is changing what software does. Three commercial decisions follow.

I

Pricing changes

Traditional software is priced by user or licence. When software performs the work itself, charging by human seat may stop matching the value delivered. Pricing needs redesigning before customer seat counts begin to fall.

II

Authority changes

A tool waits for a person to act. Software agents can create records, contact customers, trigger payments and commit spending without waiting for anyone. Companies must decide where human approval is still required.

III

Margin changes

The strongest position may no longer belong to the company with the best interface. Value moves toward whoever controls the workflow, the rules that govern it, and the evidence of what happened.

Every layer agents need is a place someone can charge rent.

The Stakes

What it costs to get this wrong.

Pricing is the slowest thing to change and the most expensive to get wrong. A seat-based contract signed this quarter can run for three years while the customer's headcount falls underneath it.

Authority is the decision nobody owns. If nothing states what software may do on its own, the answer gets set by whoever ships first, and it tends to be discovered after something has already been committed on the company's behalf.

Position in the value chain is settled early. Once a workflow runs through someone else's rules, evidence and controls, moving it back is a renegotiation rather than a release.

The Flagship Diagnostic

The Agentic Margin Audit

Four weeks to map where agents change pricing, permissions and margin. A fixed-scope diagnostic for the CEO, CFO and board. The work is not to predict the future of AI. It is to identify which workflows change first, where pricing breaks, and what authority must exist before software can act.

Seven questions it answers

01  Which customer workflows will AI run first?
02  Which of your products become the backend rather than the interface?
03  Where does per-user pricing stop working, and what should replace it?
04  Where can software commit spending or make promises on your behalf?
05  Who keeps the margin: you, the model provider, the platform, the distributor or the system of record?
06  What has to be governed before software is allowed to act?
07  What are you doing in the next 90 days?

What the board receives

Board-ready memo with the verdict in the first paragraph, not buried in an appendix.
Workflow map of what becomes agent-run, in what order.
Pricing-risk map showing where seat and licence models are exposed.
Permission and control recommendations: authority, spend caps, escalation, audit trails, revocation.
GTM implications and a 90-day execution plan, presented in a working session with the leadership team.
The Framework

What is software allowed to do in your business?

Before software can act for your company, five things have to be settled. Who is it. What is it allowed to do. What are the limits. How do you prove what happened. How do you stop it.

Those five controls are what I call the Permission Gate, and every engagement is built on them. Companies that settle these early tend to keep control of the workflow, the customer relationship and the margin. Companies that leave them open usually find someone else has decided for them.

The Permission Gate
SOFTWARE HUMANS USE Pricing: the seat Spend: bounded by hesitation A human pauses before spending. SOFTWARE THAT DOES THE WORK Pricing: the work Spend: initiated by the software Software does not pause. THE PERMISSION GATE Identity Authority Caps & approvals Audit & evidence Revocation value moves from selling access to doing the work The margin ends up with whoever owns the gate.
SOFTWARE HUMANS USE Priced by the seat A person pauses before spending value moves from selling access to doing the work THE PERMISSION GATE Identity Authority Caps & approvals Audit & evidence Revocation SOFTWARE THAT DOES THE WORK Priced by the work Software can initiate spending The margin ends up with whoever owns the gate.

The Permission Gate

Five controls decide whether software may act: identity, authority, caps and approvals, audit and evidence, revocation. Every engagement is built on them.

Traditional software sells access to people. When software executes the work, commercial value moves toward whoever controls its identity, authority, limits, evidence and ability to be stopped.

I help software companies, boards and investors decide which work AI should perform, how to charge for it, and what rules must exist before software is allowed to act, then turn those decisions into pricing, controls and a practical 90-day execution plan.

Observe→Assist→Govern→Delegate→Scale

Nothing is delegated that has not first been governed. The audit locates you on this line and sequences the move.

Where This Lands First

The centre of gravity is capital markets and fintech. The question is arriving everywhere.

Fintech and payments

Agents need identity, spend authority, settlement and compliance before they can transact. The question: who grants authority, which flows need new rails, and where the value ends up. This is home turf: brokerage, trading infrastructure, payments and market data.

ERP and vertical software

ERP remains the system of record; agents become the operating surface. The commercial question: what is priced by user, what is priced by work, and what must be governed before software can create vendors, purchase orders or invoices.

PE-backed software portfolios

Agents will expose which portfolio companies still price by access while value moves to work. The question: where seat models are exposed, which companies face usage-cost leakage, and which need a control layer first.

Engagements start with a conversation, not a form.

Book a working session →