Your ICP was written for humans. An agent will run it anyway.
Every company buying AI sales agents this year is about to hand its targeting, its messaging and its brand reputation to software that follows instructions exactly. The instructions are the ICP. Almost nobody has checked whether theirs is still true.
What changed is not the ICP. It is what happens to a bad one.
A stale ICP is not new. Every company has one, because the market moves faster than the document does. What is new is the machine you are about to point at it.
Works around it
Sees the title is wrong, checks LinkedIn, notices the company was acquired, skips the account, and quietly tells nobody. The bad list costs you some dials and a slow quarter.
Executes it
Runs the context it is given, at full volume, without judgment or hesitation. The bad list costs you the addressable market, your sending domains and your name in the inbox, in weeks rather than quarters.
This is the part the AI tooling conversation keeps skipping. Everyone is comparing agents. Almost nobody is auditing the instructions those agents will follow. Scale is not a strategy: it is a multiplier, and it multiplies whatever you already had.
An ICP has two layers. Only one of them ages.
This is why "we already have an ICP" and "our ICP is out of date" are usually both true at the same time. The thinking holds up. The operational detail underneath it does not.
Usually still right
The problem you solve, the kind of company that has it badly, and the economics of why deals close. This tends to hold for years, and you probably do not need to pay anyone to rewrite it.
Ages every month
Named accounts, named contacts, buying triggers, channel performance and the claims in your messaging. This is the layer an agent actually consumes, and it is the layer nobody re-checks.
Published estimates of how fast B2B contact data decays range from roughly twelve to thirty percent a year, and almost all of them come from companies that sell data. We do not quote an industry average at you. We measure yours.
What we measure
Five drift dimensions, weighted to one hundred points. The first two are data checks. The last three test your ICP against your own recent wins and your own outreach results, which is where most of the surprises are.
| Dimension | Weight | The question it answers |
|---|---|---|
| Account fit drift | 25 | Do your target accounts still match your own criteria? Who was acquired, closed, shrank or changed their stack? And which accounts fit you today but are missing from the list entirely? |
| Contact decay | 20 | Are the named people still at the company, and still in the role you recorded? |
| Signal relevance | 20 | Do the buying triggers in your ICP actually appear before your recent closed-won deals? What signals show up that your ICP does not track? |
| Message and channel fit | 20 | What has happened to reply and meeting rates by sequence over time, and do your core claims still match what buyers say they care about? |
| Buying committee drift | 15 | Are your ICP personas the titles that actually showed up in the deals you won? |
Sample sizes appear on every number in the report. We do not claim more precision than the sample supports.
The Context Score
One number, four bands, and a verdict you can act on the same week.
Roughly where would you land?
Four questions, no email required, nothing stored. This is an indication rather than a score. The real one comes from your data, not from four buttons.
Estimate your Context Score
Answer honestly. The useful version of this exercise is the uncomfortable one.
This estimate uses four proxies. The real audit scores five weighted dimensions against your account list, your CRM exports and your sequence data, and shows the evidence behind every finding.
What you get
A readout your CRO can forward to the CEO, not a binder nobody opens.
- The Context Score with the breakdown across all five dimensions, so you can see which layer is failing.
- Five to seven drift findings, each with the evidence attached. Not "your data is old" but "38 of 120 named VPs have left or changed roles."
- The cost of staleness, estimated in your own numbers at your current outreach volume: wasted sends, wasted rep hours, and accounts you never touched.
- A whitespace list of 25 accounts that fit you today and are not on your list. This is the free sample of what a rebuilt ICP produces.
- A keep, fix, rebuild map that separates the durable core from the perishable layer, so you do not pay to redo work that is still right.
- An agent readiness verdict: whether your context is safe to automate, and what has to change first if it is not.
Format: a six to eight page PDF and a sixty minute readout. The scored data goes to you as a spreadsheet you keep.
How it runs
Ten business days from the moment we have the data. About two hours of your team's time in total, which is the whole point: you should not have to run a project to find out whether you have a problem.
What you send us
- Your current ICP or go-to-market document, in whatever form it exists.
- Your target account list and a contact export. We sample up to 200 accounts and 300 contacts.
- A CRM export of closed-won and closed-lost from the last 12 to 24 months, with contact roles.
- Sequence or campaign performance, with reply and meeting rates by sequence.
- Your current outbound messaging: templates, one-pagers, and how you position on your own site.
What we do with it
| Days | Step |
|---|---|
| Day 0 | Kickoff call, data request, NDA signed before anything moves. |
| Days 1 to 3 | Enrich and verify the account and contact samples. |
| Days 3 to 6 | Test your triggers and personas against your closed-won record, and analyze the sequence data. |
| Days 6 to 8 | Message review, scoring, and the draft readout. |
| Day 9 | Internal review of every score and claim before you see it. |
| Day 10 | Sixty minute readout, with the recommendation attached. |
What it costs
Where it sits
| Step | What it is | Price |
|---|---|---|
| Context Check | 25 accounts, account and contact drift only, one page back to you. | Free |
| Context Lab | The full five dimension audit, the score, the whitespace list and the readout. | $3,500 |
| Phase One | ICP, Message, Channel and Signal Labs. Rebuilds the layer your agents run on. | $14k to $22k |
| ARC | Agent-run outbound, on a context that has been tested rather than assumed. | Retainer |
You are not obliged to take the next step, and plenty of companies should not. If the score comes back Fresh, we will tell you that and you will have spent $3,500 to find out you were right.
The objections we actually get
Find out before the agent does.
Thirty minutes, no deck. We will look at what your ICP is built on and tell you whether it is worth auditing at all. If it is fresh, that is a short call and a free answer.