Ready for a controlled pilot
The assessed data supports testing the defined job, with documented limitations and checks.
AI Agent Data Readiness Assessment
Find out whether your data can support an AI agent before you pay to build one.
You have a job you want an agent to do. I assess the records, rules and permissions behind that workflow with the people who do the work today, so you can decide what is ready to test and what needs fixing first.
Led by Ryan Ginsberg · Delivered through Unified Support Solutions
Start with a specific job
A workflow your team needs to trust.
For the person responsible for a workflow in HubSpot or connected business systems who is considering an agent, or whose pilot produces answers the team cannot confidently use.
Bring a workflow of your own. We first agree what it needs to accomplish, including whether an agent is a useful way to do it.
Prepare for an account meeting
Connect the right company, contacts, deals, notes and open issues. Check whether the context is current and belongs to the right account.
Prioritize and route a ticket
Apply customer priorities, service commitments and assignment rules. Check whether those rules are documented and the necessary ticket fields are present.
Identify a renewal that needs attention
Connect contracts, renewal dates, billing status and account ownership. Find where the records disagree before automating follow-up.
Review a quote or discount
Compare products, prices and discounts with the approved rules. Distinguish an allowed price from an exception that needs human approval.
A decision for each workflow
Know the next step before committing to a build.
Fix the data first
Specific gaps would prevent a useful test. You receive the priorities and recommended corrections.
Choose a different starting point
The workflow is not practical with the data and access available today. I explain the constraint and identify a narrower option where the evidence supports one.
The recommendation concerns the assessed data and workflow. The agent itself still needs testing before production use. Where evidence is insufficient, the finding stays unverified.
What I assess
The records, the rules and a way to check the answer.
We trace real records and review past cases with the people who know the work. Missing access, uncertain answers and evidence gaps stay visible.
Records you can rely on
Completeness, accuracy, freshness, field meanings, duplicates and relationships. Which system owns each value, and whether integrations overwrite one another.
Rules the agent can follow
Knowledge that lives in people’s heads, permitted reads and changes, human approvals and a named owner for the workflow.
A result your team can verify
Expected answers, difficult exceptions and cases that should return to a person. A past outcome becomes a reference answer only after your owner reviews it.
What you keep
A decision, with the work behind it.
Your team can use the deliverables with me, another implementation partner or your own staff.
- A decision brief
- A recommendation per workflow, with evidence, limitations and unresolved questions.
- A prioritized gap list
- Business consequences, recommended owners and estimated effort to address each gap.
- A workflow data dictionary
- The fields, relationships, meanings and source systems in scope.
- An access and approval specification
- Proposed reads, actions, human review and ongoing ownership.
- An initial evaluation set
- Agreed cases and expected results reviewed with your owner, including exceptions and unresolved cases.
- Repeatable profiling checks
- The checks behind quantitative findings, with instructions and access requirements to rerun them. Interview findings retain their supporting notes.
Scope and process
Agree the job.
Inspect the evidence.
One to three related workflows, with the systems, record types, data period, samples and deliverables agreed before starting.
How the assessment runs
- Define the job, intended actions and what a correct result looks like.
- Profile the agreed data and document repeatable checks.
- Trace real records through normal cases and exceptions.
- Interview the people who own and perform the work.
- Review the recommendation and priorities with your team.
What I need from your team
Read-only access or suitable exports, existing documentation and relevant past cases. Plan for approximately two to four hours of interviews across the workflow owner, your admin and the people doing the work.
Access setup and review of expected answers may require additional time. Where reliable reference answers are missing, that becomes a documented gap.
Fee and timing
The fee is fixed in the proposal before work begins. Plan for two to three weeks once the required access, materials and participants are available. Scope changes are agreed before additional cost.
Where the assessment ends
The assessment is read-only. Data remediation, agent development and production testing are separately scoped. Proposed permissions are documented here; enforcement needs verification during the pilot.
This is a specific Foundation Audit. It can stand on its own and is optional when the necessary investigation is already included in an agreed implementation project.
Bring the workflow you have in mind.
In a free 30-minute introduction, we discuss the job, the systems involved and whether an assessment is a useful next step. Bring one ordinary example and one exception that is difficult to handle today.
Fit and scope first. The paid assessment is agreed separately.
Situation to discuss