Agent Readiness Assessment

The six dimensions

Before we write a line of agent code for anyone, we score their business on six things. This is that instrument written out in full: what each dimension measures, why it matters, what good looks like, and what to fix first.

Take the assessment Download as PDF

How to use this

It's meant to be used, not just read. Run it on your own organization in a room with the people who would own the result. You don't need us there.

Place your organization on each row of the matrix.

Be honest rather than aspirational. All the value is in where it disagrees with the story you tell outside the company.

Expect an uneven profile.

Nearly every organization is agent-ready on two or three dimensions and absent on one. The absent one is the finding, and it is usually the one nobody owns.

Sequence the work. Do not fix everything at once.

The six are not equally weighted and they are not independent. The reading order is at the bottom of this page.

The same six dimensions with the site chrome stripped, for circulating internally.

The maturity matrix

Six dimensions, four levels. Most organizations land in more than one column. That's the point: the shape of the profile tells you where to start.

Agent readiness maturity by dimension. Go Fig, 2026.
Dimension	Absent No mechanism at all	Manual Happens only because a person holds it together	Assisted Partly systematic, real gaps remain	Agent-ready Automated, observable, and trusted
01 Systems connectedness	Scattered and nothing talks to anything	Many, stitched together by hand	Several, partly connected	Connected, syncing automatically
02 Manual work	It eats the week and the real work happens in the gaps	A day or more a week per person	A few hours a week per person	Very little, most of it runs on its own
03 Reporting you trust	We decide on gut feel because the numbers come too late	Most of a week, and people argue about whose number is right	A day, someone pulls it together	Minutes, from a live dashboard everyone trusts
04 Follow-through	Things fall through and we find out from the customer	A few days, if someone remembers	Usually the same day	Same day, without anyone chasing
05 Knowledge capture	When someone leaves, what they knew leaves with them	It lives in people's inboxes and heads	Mostly written down, hard to find	Captured automatically and searchable
06 Agent readiness and traceability	No way to tell, so we do not automate at all	We would see the result, never the reasoning	We could reconstruct most of it with some digging	Every action logged and approved before it runs

The dimensions in full

Systems connectedness

How many systems hold your business data, and do they talk to each other?

What it measures

How many systems hold a fact about your customers, your money, and your work, and whether those facts reconcile without someone doing it. Count the CRM, the accounting system, the ERP or PSA, email and calendar, call notes, and the spreadsheets in between.

Why it matters

An agent can only reason over what it can see, and so can a report. Point either at one system and it inherits that system's blind spots. A follow-up goes to a customer whose invoice is 60 days late, or a margin report misses the costs that live in the ERP. When an agent looks unreliable in production, the model is rarely the cause. Usually it's working from half the picture and being right about the wrong thing.

What good looks like

Every system lands in one private place on a schedule you control. That place is infrastructure you own rather than a vendor's copy of your business.

What to fix first

Inventory before integration. List every system that holds a fact you report on or act on, and mark which ones can share data with other tools (an API) today. The gap list is your integration backlog. It's almost always shorter than the room expects.

Manual work

How much of your team's week goes to copying data between systems, building reports, and chasing updates by hand?

What it measures

The time people spend retyping between systems, rebuilding the same report, logging calls, and chasing who did what, against the work they were hired for. Measured per person, per week.

Why it matters

This is the dimension that funds the program. Hours lost to manual work turn straight into a dollar figure the budget owner can approve. If you can't state the number, you can't make the budget case, and the project stalls at the budget conversation rather than the technical one.

What good looks like

The report refreshes itself, the CRM note is drafted before anyone reopens their laptop, and nobody retypes into one system what another system already knows.

What to fix first

Measure it before you automate it. Take a two-week sample across one team and count honestly. The baseline is what your pilot gets judged against. A pilot without a baseline gets judged on vibes.

Reporting you trust

When leadership asks how the business is doing, how long until there is a number everyone trusts?

What it measures

How long it takes to answer a basic question about revenue, margin, cash, pipeline, or capacity, and whether everyone in the room agrees with the answer.

Why it matters

Agents act on the same numbers leadership reads. If the numbers need a person to reconcile them before anyone trusts them, an agent acting on them will be confidently wrong. Reporting is also where trust in the whole program is won or lost. A dashboard that ties to the books earns the right to put an agent on top of it.

What good looks like

The numbers leadership asks for most are on a live dashboard, every figure opens the record behind it, and a repeat question becomes a scheduled report.

What to fix first

Pick the five numbers leadership asks for every week and trace each one to the system it comes from. Where two systems disagree, that disagreement is the first thing to fix.

Follow-through

When work passes from one person to the next, how reliably does the next step happen?

What it measures

Elapsed time from a handoff moment to the step that follows it: a won deal reaching the team that delivers it, a proposal getting its signature, a customer email getting its reply, an invoice going out.

Why it matters

This is where agents protect revenue rather than save time. Most work is lost between people, not inside a task. Every other dimension gives hours back. This one saves the client who waited a week for a kickoff and the invoice that went out a month late. It also has the cleanest before-and-after measurement of the six, which makes it the pilot most likely to survive scrutiny.

What good looks like

The handoff is drafted the moment it is due. A person approves it until the team trusts it enough to stop reading every one.

What to fix first

Measure it. Most companies have never measured how long a won deal takes to reach delivery, or a proposal to come back signed. The measurement alone changes behavior before a single agent ships.

Knowledge capture

Does what your team learns in email, calls, and documents end up somewhere the business can find it?

What it measures

Whether what was said on a call, agreed in an email thread, or written in a document reaches a place the business can search, next to the systems of record.

Why it matters

Most of what a company knows never reaches a system of record. It sits in inboxes, call recordings, and documents. An agent that can't read it drafts from the CRM field instead of from what the customer actually said, and a question to an AI analyst gets half an answer.

What good looks like

Email, call transcripts, and documents sit beside the structured data, so an answer can draw on both and a draft carries what was actually agreed.

What to fix first

Start with one source: the call notes or the shared inbox for your most important customers. Capture it next to the CRM and check whether the answers get better.

Agent readiness and traceability

If an automated action ran on your behalf, could you tell what it did, what data it read, and why?

What it measures

Whether you can reconstruct what an automated action did, which records it read to decide, and who approved it. And whether a person could have stopped it first.

Why it matters

This is the dimension that decides whether the program ships at all. The other five determine whether agents will work. This one determines whether you will be permitted to run them. Security review, legal, and the executive who owns the customer relationship all ask the same question, and there is no version of the answer that starts with we would see the result. Programs that put this off build something impressive in the pilot. Then they die at the readiness review with their political capital spent and nothing ready to ship.

What good looks like

Every read, action, and approval is logged. The business owner decides, per agent and per action, whether it runs on its own or waits for approval. Changing that takes a click, not a project.

What to fix first

Decide on paper, before the first agent, when an action gets to run on its own. Everything starts out waiting for your OK. The agent drafts, a person approves. Then decide what earns an action the right to run on its own: a number of clean approvals, a rejection rate under a set limit, or sign-off from a named owner. Writing this down early is free. Retrofitting it is not.

Where to start when your profile is uneven

The six dimensions do different jobs. Treating them as a checklist to finish in order is the most common way this exercise gets wasted.

Foundations

Systems connectedness and traceability

These two gate everything else. One decides what an agent can see, the other decides whether you are allowed to let it act. Neither produces value on its own. That's exactly why both get put off. And putting them off is what kills agent programs at the readiness review, not at the demo.

Fastest return

Manual work and reporting you trust

These two pay for the program. They turn into hours, and hours turn into a dollar figure your budget owner will fund. Trusted reporting also improves the data the agents run on, so it compounds.

Upside

Follow-through and knowledge capture

These two protect revenue and make every answer better. That's a stronger business case but harder to attribute cleanly. Run them once you have the credibility from the first two, and measure the baseline before you start.

The mistake worth naming

Teams try to finish systems connectedness before starting anything. You do not need every system connected. You need the systems your first agent reads. That's usually two or three, plus a written governance policy. A narrow agent running against complete data for one workflow beats a broad one running against partial data for ten. It also produces the evidence that funds the rest of the program.

Want the score instead?

The assessment turns the six dimensions into six questions. In about two minutes you get your score, the hours and follow-through risk behind it, and what to fix first.

Take the assessment

See which of these Go Fig would run for you first

A solutions engineer walks through your systems with you and maps which work the agents can take off your team. You approve every action.

Schedule your free evaluation