cost saved
above plan
€240k – €380k
Across 7 documented use cases. Net of AI tool spend (build + run).
talon · get a grip on ai adoption
The quarterly AI report your board actually wants.
One page a quarter: what AI is running across your company, who owns it, what it can do without a person, and what it produced.
Get the sample report →The methodology is open-source. AI is new. The reporting discipline is not.
the problem
Pilot purgatory has a price tag.
Most executive teams have now sat through some kind of AI fluency assessment. The score sits in a folder. Nothing changed. Meanwhile 63%1 of companies report no measurable earnings impact from AI at all.
Your board does not want to know your fluency level or how many seats you bought. They want financial impact. Cost saved. Speed gained. Revenue earned. In euros, with assumptions a CFO can sign off on.
Scoring is a snapshot. Activity dashboards are noise. Reporting is an operating system. That is what Talon runs.
the exposure
Everybody is turning on agents. Almost nobody can list them.
Personio screens CVs. HubSpot works prospects. Intercom answers customers. Someone in finance built one to review invoices. These get switched on inside tools you already pay for, by people who had no reason to tell IT.
So most companies can't say what AI is running, what data it reaches, or what it can do on its own. None of this is unusual. There was simply never a moment when the decision had to be written down.
Giving people access to AI is easy. Knowing what they turned on is the governance problem.
| agent | owner | can act | data |
|---|---|---|---|
| CV Screening Agent HR · Personio | People | No | Applicant |
| Prospect Research Agent Sales · HubSpot | Sales | Yes | CRM |
| Customer Reply Agent Support · Intercom | Support | Yes | Customer |
| Invoice Review Agent Finance · built in-house | no owner | Yes | Financial |
the report
One page. Three numbers. One conclusion.
Cost, speed and revenue for the quarter, plus the exposure page your legal team will ask for.
speed gained
on track
31 – 48%
Cycle time reduction on three workflows, weighted by volume. Per function.
revenue impact
in measurement
€380k – €620k
Attributed via three documented methods. CFO signed off.
behind the numbers ↓
Revenue grew 18% on flat engineering headcount through Q1. Capacity gained, not headcount added.
Every use case carries what it can do and what it produced. Same row, same quarter.
| What it is, and what it can do | What it produces | ||||
|---|---|---|---|---|---|
| ai use case | owner | can act | data | outcome | status |
| Intercom Fin | Support | Yes | Customer | First response 4.2hr → 11min | verified |
| BrightHire | People | No | Applicant | Interview-to-offer 21 → 12 days | review due |
| CRM prospecting | Sales | Yes | CRM | Reply rate 4.1% → 7.2% | attention |
| Candis | Finance | With approval | Financial | Invoice processing 14 → 4 min | verified |
| Claude Code | Engineering | Yes | Code | Defect escape ratio 12% → 7% | verified |
how it works
Three steps. No fluff.
- 1.
Each function says what it uses AI for
In their own words, in a few minutes. No scanning, no integration project. The AI that matters is switched on inside tools you already bought, and no discovery tool can see it. So Talon asks the people who turned it on.
- 2.
Report quarterly.
Cost, speed, revenue and capacity gained, rolled up by tier, by function, by use case. Every claim carries a documented assumption and a sample size, so a CFO can argue with the number instead of dismissing it.
- 3.
Track change over time.
Drift, progress, killed pilots, scaled wins. Last quarter beside this quarter, so the board sees a trend rather than a snapshot, and so a claim that quietly stopped being true shows up.
the methodology
Open-source. Run it yourself.
Talon is built on the AI Adoption Playbook, an open-source framework that runs inside Claude Code. Two stages, twelve skills. Stage 1 diagnoses where AI adoption is stuck. Stage 2 measures whether you can defend the value to a board. Same methodology you see in the sample report. Use it yourself, or talk to me about helping you implement it.
View the playbook on GitHub →what talon tracks
The metrics that matter to a board. Not the ones that don't.
One view. Three questions.
one
What are we using
- AI use cases by function and tier
- Who owns each one
- Production versus pilot
- Time to first value
- Pilots scaled and pilots stopped
two
What needs attention
- What can act without a person approving it
- What reaches personal, employee or applicant data
- What has no accountable owner
- EU AI Act high-risk classification and audit readiness
- Reviews due and reviews overdue
three
Is it working
- Cost saved net of total AI ownership
- Speed gained by workflow
- Revenue impact with a named attribution model
- Capacity gained on a flat FTE base
the sample
Download the sample report.
A four-page PDF showing what your next board update could look like. Fictional company, real methodology.