AI business value framework
Describe the proposal in plain English, the engine will separate what you supplied from what it estimated, then return a provisional Accelerate, Fix, or Stop direction.
The first assessment starts with disclosed AI BVF planning assumptions, so it tells you what evidence is missing before the value range can become a budget case.
Return in this browser profile to review the evidence and re-score. Records are stored locally; clearing browser data removes them.
Use organisation-neutral wording and leave out confidential information. Assessment content stays in this browser, while anonymous journey events record whether the assessment reached a verdict and returned for re-scoring.
Choose the closest position in each pull-down, use Not known where the business has not checked it.
Current direction
These four selections update the direction, they do not claim that supporting evidence has been reviewed.
Use the next material question when an accountable owner can name the source and review point behind the answer.
Each standard must clear its own threshold, while a funding decision still needs the source material and review point confirmed by the accountable owner.
Does it move a named board outcome with a baseline, target and accountable executive?
Do measured volume, full cost and a finance-tested capture rate support the case?
Have the workflow, roles and measures changed, with a sponsor and product owner able to deliver them?
Are decision rights, human override, monitoring, incident response and withdrawal authority tested?
What the answer means
Estimated inputs stay visible, a weak case gets Fix, and a binding risk can stop the work before more money follows it.
Accelerate means the thresholds clear on the supplied inputs; accountable owners still validate the funding case. Fix means close the named gaps and re-score. Stop means the current case should not proceed.
Each verdict carries the resolved inputs, modelled value range, decision score, rules and audit record from the same open engine.
Continue in Claude
The connector can test the evidence, build the improvement plan and score the initiative again when the work changes.
In Claude, open Settings, then Connectors, choose Add custom connector and enter:
https://mcp.aibvf.com/api/mcpThe hosted connector runs on the AI BVF server and provides 13 tools. Share only information you are authorised to send. See hosted and local setup instructions.
Use the result link above or describe the initiative again, then ask Claude to use assess_ai_initiative and test the evidence behind the verdict.
Return when the workflow, roles, decision rights, measures or business case have changed, and see whether the initiative has earned a different verdict.
https://mcp.aibvf.com/api/mcp as an MCP server, or run npx -y aibvf-mcp locally.