AI Readiness Assessment
Two to four weeks inside your business. It ends with a written plan, and an honest list of the things we think you shouldn't build.
- Duration
- 2–4 weeks
- Fee
- $10,000
- Your time
- 5–8 interviews, one readout
- Output
- Written action plan
What this is
Most companies that try AI start with a tool and go looking for something to point it at. This runs the other direction.
We spend two to four weeks talking to the people who do the work and looking at the systems they do it in. We come out with a ranked list of work that can actually be handed to an AI agent today, a written job description for the top candidates, and a plain account of what stands in the way of the rest.
Some of what you want to automate won't survive the filter. That's the part most of our clients say they got the most out of, not because it's satisfying to be told no, but because they stop spending money on it.
The filter that comes first
Before any framework, two questions. Both have to be yes or the workflow is out.
- Is the task hireable? Could you hand it to a capable intern with written instructions and have it done correctly? If it depends on judgment nobody has written down, it isn't ready.
- Does the data exist somewhere? A spreadsheet, a database, an API, an export: anywhere an agent can be pointed at it. What it can't be is in someone's head, or on sticky notes, or reconstructed from memory each time. If the data has to start being recorded before any of this works, we'll say so and tell you what the process change is.
Most failed AI projects failed here, before anyone wrote a prompt or picked a vendor. The task wasn't hireable, or the information it depended on was never written down anywhere. The model was never the problem.
The framework: DOCS
Everything that survives step zero gets written up the same way: the way you'd write up a new hire before their first day.
- Data
Can it see what's happening? Live access to the numbers, the tickets, the inventory, not a stale export somebody remembers to run.
- Orientation
Does it understand the job, how it's done here, and what counts as doing it well? Not "flag weird expenses," but what weird means at your company, the steps a person takes to check, and the standard the output gets measured against. The job description, the method, and the definition of success and failure.
- Capabilities
What is it allowed to do? Flag for a human, recommend, or act. You don't give a new hire signing authority on day one.
- Supervision
Who manages it, and what happens when it's wrong? Audit trails, a review cadence, a person whose name is on it, a way to turn it off.
Skip one and it doesn't fail loudly. It fails quietly, over weeks, and the conclusion everyone draws is that AI doesn't work here.
How the four weeks run
-
Before we start
Intake questionnaire
A short written questionnaire, filled out by whoever knows the business best. What systems you run, where the friction is, who touches what. It takes about twenty minutes and it means the first interview starts at the second question instead of the first.
-
Week 1
Interviews
Forty-five minutes each with five to eight people: leadership, plus the people actually doing the work. We ask what takes too long, what gets re-done, what nobody chases until it's late, and what report someone rebuilds by hand every month.
-
Weeks 1–2
Systems review
Read-only access or screenshares of the systems involved. What data is in them, what shape it's in, who owns it, and how an agent would reach it. This is the unglamorous half of the engagement and it's where most of the real answers come from.
-
Weeks 2–3
Filtering and drafting
Every candidate workflow goes through step zero. The ones that pass get written up under DOCS. The ones that fail get a one-line reason and a note on what would have to change for them to pass later.
-
Weeks 3–4
Readout
Ninety minutes with your leadership team. We walk the plan, you argue with it, we change what's wrong. You keep the document either way.
What you get
-
Opportunity inventory
Every workflow anyone raised, ranked, each marked pass or fail against step zero with the reason in one line.
-
Systems map
Where your data actually lives, how an agent would reach it, and which parts of the work are still living in people's heads rather than in a system. Where a process change has to come first, it's named.
-
Job descriptions for the top candidates
Two or three, written in full: what the agent is responsible for, what data it can and can't see, what good looks like with edge cases, its authority level, the escalation path, and the review cadence. Detailed enough to hand to an engineer or to a vendor.
-
Technical action plan
The build-out your engineering team can execute from, written for them rather than for you: where the data gets consolidated and in what shape, which harness the agents run in, which connections each system needs and whether one already exists or has to be written, how credentials and permissions get scoped, and what logging has to be in place before anything runs unattended. Every recommendation comes with the reasoning, so your team can disagree with it on the merits.
-
Sequencing and level of effort
What to build first and why, rough effort for each, and what it needs from your team, including the data cleanup nobody wants to hear about.
-
The not-yet list
What we recommend you don't build, and the specific condition that would change our answer.
What this isn't
- We don't build anything during the assessment. No pilot, no prototype, no code. If you want something built, that's a separate conversation after you've read the plan.
- We won't put an hours-saved number in the document unless we can tie it to something your team already measures.
- The assessment may conclude that one workflow is worth doing right now. That's a real and fairly common outcome.
- The technical plan assumes someone will execute it: your team, an existing partner, or us. If you don't have anyone to hand it to, that's worth knowing before you start rather than after.
Who runs it
Jason Raede
CEO, Pliable · jason@pliable.co · LinkedIn
Jason was CTO at multiple companies before founding Pliable, and has since built hundreds of agentic workflows inside small and mid-market companies, the kind with real revenue and no engineering team. Both fully autonomous and human-in-the-loop, in production. Not demos.
He runs the interviews himself. There is no junior team behind this.
- CEO of Pliable, a data platform built for mid-market companies without engineering teams.
- Advises KeyBanc Capital Markets on AI market trends.
- Has shipped agentic workflows (expense review, invoice matching, ticket triage) across dozens of mid-market operations.
"I'd rather tell you three of your five ideas won't work than sell you four of them."
Jason Raede, Pliable
Starting one
Fill out the questionnaire. It takes about twenty minutes, and it's enough for us to tell you whether you have a problem worth this exercise at all. If you don't, we'll say so before anyone signs anything.
The assessment is $10,000, fixed. It buys the plan, not an obligation to build anything with us.
Pliable handles the first two components of DOCS: Data and Orientation. If you move forward with Pliable after the assessment, the full $10,000 is credited against your first year.