Resolon Start a conversation

InsightsAI funding

Published Updated 4 min read

Four questions a board asks before it funds AI

Most AI programs stall between what leadership asks for and what engineers build. These are the questions that decide whether a program gets funded, and what a good answer looks like.

Chad Neptune, CEO and Chief AI Architect

Contents

A board does not fund “AI”. It funds a result.

Will this move revenue?

The strongest proposals start with one workflow, not a platform: the one that moves revenue or cost the most, and the number that will prove it moved.

Pick the measure before anything is built and record where it stands today. If a program cannot name its number up front, it cannot show a result later.

What gets built follows from that choice: the agents, data and integrations behind that one workflow, and nothing that does not serve it.

For leaders Ask for the one workflow and its number before you read anything about technology.

A presenter stands beside a flipchart and a wall screen while colleagues listen in a meeting room

What will it cost, and what can go wrong?

AI costs move with use. Every call to a model has a price, and usage grows when a workflow works. A finance team needs three things before it signs: the cost to serve each unit of work, a spending ceiling the system cannot cross, and a risk register that names what can go wrong and who owns each risk.

On the engineering side that becomes model routing (the least costly model that does the task well), per-call budgets, guardrails on what the system may do, and a fallback for every dependency that can fail.

For leaders Do not sign a budget that has no ceiling and no named owner for each risk.

Taking an AI program to your board? Talk to a principal.

Can a regulator audit it?

In a regulated business the question is not only what the AI decided, but who approved it and whether you can show it. Decision rights come first: what the AI decides on its own, what it only recommends, and what a named person must approve.

Those rights become approval gates in the workflow, an audit trail of every action, access limited by role, and evaluations that test the system against known answers before and after each change.

For leaders Decide in writing what the AI may do alone, what it may only recommend, and what a named person must approve.

When does it ship, and will people use it?

A program that ships and is not used has still failed.

Plan the rollout in phases, with an adoption target at every stage, not just a launch date.

New workflows should run alongside the old way of working until the people who own that work sign off. Every release needs a plan, testing, monitoring and a way to roll back.

For leaders Judge the program on whether people use it, not on whether it launched.

The pattern behind the answers

Problem before platform. Trade-offs in the open. Change carried through to adoption.

A program built around these four questions arrives at the board with its answers already in hand.

Without an answerWith an answer
Will this move revenue?A platform, with no number to prove itOne workflow, and a measure recorded before anything is built
What will it cost, and what can go wrong?Costs that grow with use and have no limitThe cost to serve each unit of work, a spending ceiling, and a risk register with an owner for each risk
Can a regulator audit it?An AI decision nobody can trace to a personDecision rights, approval gates, an audit trail and access limited by role
When does it ship, and will people use it?A launch dateA phased rollout with an adoption target at each stage, run beside the old way until the owners sign off
Figure 1. The four questions, and what a funded proposal answers

For leaders Take Figure 1 into the meeting and ask which column your proposal is in.

What to do this quarter

Before the next AI proposal reaches your board, put the four questions to its sponsor and ask for the answers in writing.

  1. Name the one workflow and its number.

    Write down where the measure stands today, before anything is built.

  2. Set a ceiling on spend.

    Ask for the cost to serve each unit of work and a limit the system cannot cross.

  3. Decide who approves what.

    List what the AI decides on its own, what it only recommends, and what a named person must approve.

  4. Plan adoption as part of the launch.

    A phased rollout, an adoption target at each stage, and the old way running alongside until the owners sign off.

Frequently asked questions

What do boards want to see before they fund an AI program?

A result they can measure. The strongest proposals start with one workflow rather than a platform, name the number that will prove it moved, and record where that number stands today.

Do we need an AI platform before we start?

No. Start with the one workflow that moves revenue or cost the most. What gets built is the agents, data and integrations behind that workflow, and nothing that does not serve it.

How should we budget for AI costs that move with use?

Budget for the cost to serve each unit of work, not for a flat licence. Set a spending ceiling the system cannot cross, send each task to the least costly model that does it well, and keep a risk register with a named owner for each risk.

How do we make an AI system auditable?

Decide first what the AI decides on its own, what it only recommends, and what a named person must approve. Then build those decision rights into the workflow as approval gates, with an audit trail of every action and access limited by role.

Why do AI programs fail after they ship?

A common reason is that people do not use them. Plan the rollout in phases, set an adoption target at every stage, and run the new workflow alongside the old way until the people who own the work sign off.

Share this article

New each month. Follow by RSS

Senior-led, start to finish.

A senior Resolon lead is accountable for every engagement, from the first conversation to handover, with a team built for the work. For an active engagement or a new requirement, start a conversation.