AI Readiness Assessment: A Checklist for Service Firms

An AI readiness assessment checks one workflow: usable data, defined work, people who can review, and clear rules for the information. Record what is known, what is missing, and who will close each gap. For a short firm snapshot, take the AI readiness score for service firms.

Which workflow are you assessing?

Choose one recurring preparation job, such as assembling a client file. Write where it starts, what a reviewer should receive, and what stays with the professional. If you have not chosen yet, start with where to begin AI adoption.

Use that same workflow in every section. “Our data is good” does not say whether this job’s documents are available.

Is the data usable and authorized?

• Inputs. List the documents, records, and approved firm knowledge the job needs.
• Authority. Name where each input lives and who can authorize its use.
• Quality. Check that it is current, readable, and tied to the right client.
• Gaps. Record missing, conflicting, or duplicate information. Do not assume the agent will fix it.

Keep one complete example and one incomplete example. The AI data pipeline guide covers connecting and checking that information.

Can your systems support the workflow?

• Products. Record the exact products and versions involved.
• Read. Confirm an export, a supported connection, or another agreed route.
• Write. Check write access separately from read access.
• Owner. Name who will test access and resolve integration constraints.

Precision’s AI agents service treats accessible data and integration feasibility as prerequisites. A product name does not establish what your installation permits.

Are the workflow and expected result clear?

• Steps. Write the criteria staff use to prepare the work today.
• Judgment. Separate repeatable preparation from professional decisions.
• Output. Define what the reviewer receives, and the evidence they need.
• Stops. Name the exceptions that must wait for a person.
• Baseline. Include preparation, review, and corrections. Do not substitute an assumed time saving.

Do people have ownership and review capacity?

• Owner. Name a process owner and the people who do the work.
• Review. Identify who reviews, who approves, and who handles exceptions.
• Time. Schedule a test of the prepared work and a place to record corrections.
• Upkeep. Decide who owns changes when systems or procedures change.

NIST MAP 3.5 recommends documenting oversight and testing it under conditions similar to deployment. See human in the loop AI for questions to test that handoff.

Are the governance decisions documented?

• Permissions. Identify what the workflow may use, and which consents to check.
• Tools. Record approved tools and the data-handling terms that apply.
• Boundary. Confirm review before any client communication or consequential action.
• Records. Name the approval record, the retention rule, and the escalation owner.

Use your firm’s own policies. This checklist does not state a legal conclusion.

What changes for CPA firms and law firms?

The permissions and review sections carry profession rules. This is not tax or legal advice. Confirm consent wording with counsel.

CPA firms

• Consent. IRC §7216 requires the taxpayer’s written consent for uses outside the regulatory exceptions. Rev. Proc. 2013-14 keeps use consents and disclosure consents in separate documents.
• Review. The IRS says to review all AI output before relying on it.
• Billing. The same IRS guidance says billing time not spent, or double billing AI-assisted work, may violate Circular 230 §10.27.

AI for CPA and accounting firms maps the jobs to start with.

Law firms

• Consent. ABA Opinion 512 requires informed client consent before representation information goes into a self-learning AI tool. General engagement-letter language is not enough.
• Decisions. Keep opening a matter, clearing a conflict, advice and signing with the attorney.
• Time. Under Opinion 512, hourly bills reflect actual time, and learning the tool is generally not billable.

AI for law firms maps intake, conflicts and client updates.

How do you turn gaps into a first-build plan?

Mark each area confirmed, missing, or unknown. An unknown needs evidence before it becomes a yes.

• Data. Input list and sample files. A source owner checks access and completeness.
• Systems. Products, versions, and supported access. An IT contact checks read and write limits.
• Workflow. Current steps and the reviewer output. A process owner documents checks and exceptions.
• People. Reviewer duties and test time. Agree how results will be judged.
• Governance. Policies, permissions, and review records. A governance owner resolves what is still open.

Precision’s AI strategy consulting covers opportunities, data readiness, risk, and a first-build plan. Bring the worksheet to a discovery call. The next step may be preparing data or clarifying the process before an agent is built.