Process inventory
We sit with the team that feels the pain and count: how often each process runs, how long it takes, how often it goes wrong, and what the error costs when it does.
Three people, two days a month, reconciling two screens — priced.
Capability
AI Readiness Audit
Two weeks that end in a written plan: which processes are worth automating, what each would cost, and what has to be fixed first. Including the answer that you should not build — which is the answer roughly one audit in four produces.
From $2,400 · 2 weeks · fixed scope, agreed in writing
- Start to written plan
- 2 weeksStart to written plan
- Fixed, whatever the verdict
- $2,400Fixed, whatever the verdict
- Integration, not assumed
- TestedIntegration, not assumed
- The plan is yours to take elsewhere
- Any vendorThe plan is yours to take elsewhere
What your team gets
What is actually delivered
Not a strategy deck. A running system, the permissions around it, and the code in your repository.
- A ranked shortlist of processes with hours and cost estimated against each
- A tested verdict on whether your systems can actually be integrated
- A written build plan you can take to any vendor, including not us
- The package counterfactual: what off-the-shelf software would cost instead
- A data and security assessment covering where each piece of data would travel
- A 60-minute walkthrough with the people who would have to live with it
What we build
What the two weeks actually cover
Six areas, each ending in something written down. There is no discovery theatre here — every item produces a page you can act on.
Integration spike
A real test against your API or a sandbox: can we read, can we write, and what will the system refuse to do? This is the part most vendors skip and most projects die on.
Discovering in week one that the ERP has no write endpoint for purchase orders.
Data and security review
Where each piece of data sits, where it would travel, and what that means under your obligations. Written plainly enough to hand to someone who has to sign it off.
A data-flow map showing exactly what would leave your network, and what would not.
The package counterfactual
Before recommending a build we name the closest off-the-shelf product, verify what it genuinely cannot do, and compare three years of cost on both sides.
Concluding that a $200/month product covers 90% of it, and saying so.
Ranked opportunity list
Every candidate scored on value, feasibility and risk, then sorted. The first item should be the one that is worth doing and can actually be done.
Six candidates, two worth building, one worth building first.
The build plan
Scope, sequence, estimate and the things that must be fixed before any of it starts — written so another firm could execute it.
A plan you could put out to tender tomorrow.
How we build it
How the audit runs, day by day
Two weeks is short enough to be worth risking and long enough to test the thing that matters. The order is deliberate: we establish cost before feasibility, because a process that is easy to automate and costs nothing is not worth automating.
- 01
Days 1–2 · Inventory
Interviews with the people doing the work, not only the people who manage it. Volume, duration, error rate and the cost of each error, counted rather than estimated.
- 02
Days 3–5 · Spike
Credentials to a sandbox, then a real read and a real write. We produce a written list of operations each system will and will not permit.
- 03
Days 6–7 · Counterfactual
The closest packaged product named and priced, its gaps verified independently from documentation or a demo rather than from your account of them.
- 04
Days 8–9 · Scoring
Each candidate scored on annual cost, technical feasibility and delivery risk, and sorted. The scoring model is shown, so you can disagree with it.
- 05
Day 10 · Plan and walkthrough
The written plan delivered and walked through with the people who would own it, including the recommendation not to proceed where that is the honest answer.
Why we test the integration instead of assuming it
The most expensive moment in an AI project is week six, when the team discovers the ERP will not accept the write the whole design depends on. By then the budget is committed and the scope has to be rewritten around a limitation that a two-day test would have found. The spike is the cheapest insurance available in this category, and it is why our build estimates hold.
- Read test
- Against live or sandbox data.
- Write test
- A real write into a test record.
- Refusal list
- What the system will not do, in writing.
- Rate limits
- What the API actually tolerates.
The package counterfactual
In most segments, packaged software already covers the standard version of the problem. Before recommending a custom build we name the closest product, verify the failing requirement independently, show the requirement is structural rather than habit, and compare three years of cost on both sides — including maintenance, hosting and dependency on us. "The package wins" is a legitimate and frequent outcome.
- Name it
- The actual product and its price.
- Verify
- The gap, from docs or a demo.
- Structural?
- What breaks if you adapt instead.
- 3-year cost
- Both sides, honestly.
What you get if we withdraw
Around one audit in four ends with a recommendation not to build. You keep everything: the process inventory, the integration findings, the scoring and the plan. It is the same deliverable, and it has saved clients considerably more than the audit cost.
Where it lands
What this looks like in your industry
Manufacturing
Deciding between an ERP module, a custom build and leaving the spreadsheet alone.
Distribution and logistics
Establishing whether the ERP can support automated replenishment before anyone budgets for it.
Institutions and membership
Comparing membership platforms against a custom portal on three years of cost.
E-commerce and retail
Finding which of six manual steps between storefront and ERP is worth automating first.
FinTech and financial services
Mapping where data would travel before any AI tooling is approved.
Healthcare and pharmacy
Assessing what can be automated without touching anything clinically sensitive.
Technology
What we build agents with, and what we connect them to
We pick the boring option unless there is a reason not to — the framework is the part most likely to be abandoned before your system is.
- Integration testing
- Postman
- Bruno
- OpenAPI
- curl
- Vendor sandboxes
- Process analysis
- Time-and-volume study
- BPMN mapping
- Pandas
- SQL against a read replica
- Data review
- Data-flow mapping
- PII inventory
- Retention review
- Access matrix
- Feasibility prototyping
- Python notebooks
- FastAPI stubs
- Model API trials
- Cost-per-call modelling
- Market comparison
- Vendor documentation
- Product demos
- Implementation partners
- Three-year TCO model
- Deliverables
- Written build plan
- Scored opportunity list
- Integration verdict
- Data-flow map
How the engagement runs
Five phases, each with something you can hold
Every phase ends in a named deliverable. You can stop after any of them and keep what has been built.
- 01Week 0
Discovery call
What is going wrong and who feels it. We confirm the audit is the right next step, or tell you it is not.
DeliverableA written next step, free - 02Days 1–2
Process inventory
Interviews with the people doing the work, and the volume and duration counted rather than estimated.
DeliverableCosted process inventory - 03Days 3–5
Integration spike
A real read and a real write against your systems, and a written list of what each will and will not permit.
DeliverableIntegration verdict and operation catalogue - 04Days 6–9
Counterfactual and scoring
Packaged alternatives named and priced, candidates scored on cost, feasibility and risk, and ranked.
DeliverablePackage comparison and scored shortlist - 05Day 10
Plan and walkthrough
The written build plan, walked through with the people who would own it — including the recommendation not to build, where that is the answer.
DeliverableBuild plan and 60-minute walkthrough
Why Devs Core
Six reasons that are checkable
Fixed price, whatever the verdict
$2,400, whether the answer is build this first or do not build at all. Our fee does not depend on recommending a project.
The integration is tested, not assumed
A real read and a real write against your systems. This is the single thing that most distinguishes an estimate that holds from one that does not.
The plan is vendor-neutral
Written so another firm could execute it. Roughly one audit in four ends with us withdrawing, and you keep the plan either way.
We name the packaged alternative
Including when it wins. Being the people who told you that is how we get the call when the package later fails.
Cost before feasibility
A process that is easy to automate but costs nothing is not worth automating. We establish what standing still costs first.
It de-risks everything after it
Every build we quote comes off the back of one of these, which is why the four-to-six-week pilot estimate is a real number rather than an opening position.
Questions
AI Readiness Audit — the questions we get asked
What is an AI readiness audit?
An AI readiness audit is a short, fixed-price assessment that establishes whether and where AI is worth applying in a business. It inventories the candidate processes and what they cost today, tests whether the relevant systems can actually be integrated, compares packaged alternatives, and produces a ranked, costed build plan — including the recommendation not to build where that is the honest conclusion.
What do we actually receive?
A written build plan, a scored and ranked shortlist of processes with hours and cost against each, a tested integration verdict listing what your systems will and will not permit, a data-flow map, a comparison against the closest packaged product on three years of cost, and a 60-minute walkthrough with the people who would own the result.
How long does it take and what does it cost?
Two weeks, $2,400 fixed. The price does not change with the verdict, and the plan is yours to take to any vendor.
What do you need from us?
Two to four hours of time from the people who do the work, read access to a sandbox or a copy of the relevant systems, and someone who can answer questions about the process. Not a data warehouse, not a strategy document, and not a cleaned dataset.
What if the audit says we should not build anything?
That happens in roughly one engagement in four, and you keep everything — the inventory, the integration findings, the scoring and the plan. Several clients have saved considerably more than the audit cost by not starting a project, and some of them came back later when the packaged product they chose stopped fitting.
Do we have to build with you afterwards?
No. The plan is written so that any competent firm could execute it. We would rather you take a plan elsewhere than commit to a build nobody has tested the assumptions of.
Is this just a sales discovery call with a price on it?
No, and the integration spike is the difference. A discovery call cannot tell you whether your ERP will accept a write; two days against a sandbox can. That test is the reason our build estimates hold, and it is not something a free call can produce.
Can you audit an AI project we have already started?
Yes, though AI Build Rescue is usually the better fit for something already built that has stalled. The audit is for deciding what to do; the rescue is for diagnosing why something did not ship and getting it live or stopping it.
Tell us which process is costing you.
Thirty minutes, no preparation, no deck. You describe what keeps going wrong and we tell you whether AI is the answer — including when it is not.
Prefer email? contact@devs-core.com