Hundreds of AI projects taught us where to look first.
About ten minutes of questionsPublic research on the company and its marketFree by invitation
Blueprint ranks a company's AI opportunities by value against effort. You answer about ten minutes of questions, we add public research on the company and its market, and your responses are scored on the method we use in our data and AI strategy engagements.
It's a first read to decide where to look. It isn't a deep dive.
Where it comes from
After hundreds of data and AI strategy engagements, many with PE sponsors and their companies, we noticed the first conversation always covers the same ground. What does the company sell, what systems does it run, how good is its data, and what has it already tried with AI? Those answers tell us where to look first.
Blueprint writes that conversation down. It asks the questions we'd ask in a first meeting, reads what's public about the company and its market, and scores your responses the way we score them in an engagement. Its job is to get a partner and a management team looking at the same picture before anyone commits to a project.
| Building data and AI systems in production | 10 years |
|---|---|
| Data and AI projects delivered | Nearly 300 |
Sunny Sky Products
DentalXChange
Elara Caring
Tungsten Automation
Robbins Brothers
Vetta
How it works
1Name the company.
Its name and website are all we need to start. If a firm sent you the link, both are filled in already, and the report carries the firm's name.
Who: You
2Answer the questions while we read up.
About ten minutes on where the data lives, the systems, the people and how AI is used today, with follow-ups written for this company. Meanwhile we read public sources on the company and its market. You check what we found and fix anything we got wrong.
Who: You and us
3Get the written report.
It opens in the browser: where the company stands on five dimensions, ten use cases ranked by value against effort, and the first moves. It stays on My reports, where you can share it or download the PDF.
Who: Us
The method behind it
The levels aren't the AI's opinion. Each answer maps to a fixed score, the way we score a company in an engagement, so the same answers always give the same levels.
AI writes the report around that score. It reads your responses and the research, draws on the patterns from our past work, and ranks the opportunities by value against effort. For each one it weighs building, buying or connecting an AI agent, because AI is a tool that fits some work and not other work.
You correct the research before the questions, and you can flag anything in the report that doesn't match what you know.
What a report shows
Every report ranks ten use cases, scores the company on five dimensions and opens on a one-page summary. The figures below come from a sample report on a fictional company.
Point at a dot, or tab to it, to see that use case.
1Do first
Retailer deduction matching
ConnectFinance and accounting
| Cost removed a year, estimate | $600k to $1.1M |
|---|---|
| Effort to first use, estimate | 8 to 12 weeks, $90k to $160k |
| Payback, estimate | 1 to 4 months |
Point at a dimension to see its five levels.
Data foundation
Where operating data lives, how it's joined, how fresh and trusted it is
- Level 1Spreadsheets and system exports; no shared source of truth
- Level 2One or two systems (ERP, CRM) hold most data; reporting by export
- Level 3A central warehouse or lake with at least two core systems loaded and refreshed daily
- Level 4Warehouse covers finance, sales and operations; definitions documented; owned by a named person
- Level 5Data products with tests and lineage; external data (retail POS, syndicated) joined
01 What we found
Kestrel Ridge Foods can start with AI in finance and accounting, retailer deduction matching, with an AR analyst checking each match.
Kestrel Ridge Foods' data and written rules are ready for AI to take on a step of the work with a person checking it. What's missing is software already running steps today, so agents can't yet work across finance and operations.
- 01 Where it standsReady for an agent with review on one workflow. Not yet ready for agents working across departments.Software running steps of the work at level 2 of 5. No AI runs a step of the work today.
- 02 Where the value is$1.1M to $2M a yearHow this is estimatedRetailer deduction matching$600k to $1.1MReplenishment from retail sales data$500k to $900kTogether$1.1M to $2MUse cases overlap, so the sum is an upper bound.Each use case's own working is on its page.Two use cases carry most of it: retailer deduction matching in finance and accounting and replenishment from retail sales data in supply chain and logistics.Estimate, on about $62.5M of revenue.
- 03 What to do firstName an owner for AI and deduction matchingOwner: CFO. Cost in the first 30 days: $10k to $25k (estimate)How this is estimatedName an owner for AI and deduction matching, days 0 to 30$10k to $25kOwner: CFO. Staff time included..
Hover or tap any figure to see how it's estimated. No figure in this report assumes anyone's role is cut.
AI wrote this report from your responses, public research and patterns from Blue Orange's past client work and case studies. Treat it as recommendations, and ask Blue Orange to verify anything you'll act on.
What research found
- Kestrel Ridge Foods makes packaged frozen meals sold through Kroger, Albertsons and Sprouts.kestrelridge.example/where-to-buy
- Kestrel Ridge Foods runs NetSuite for finance and orders and Salesforce for its retail accounts.respondent edit
- The team page lists one data analyst and no data engineer.kestrelridge.example/team
Five-dimension profile
Where the foundation stands
2.6Mean of 5
- Data foundationHigh confidenceLevel 3 of 5Typical for packaged food brands of similar revenue and headcount: level 2 of 5.A central warehouse or lake with at least two core systems loaded and refreshed daily.
- Context layerHigh confidenceLevel 3 of 5Typical for packaged food brands of similar revenue and headcount: level 2 of 5.A maintained, searchable store of documents and definitions that software can retrieve from.
- Agentic readinessHigh confidenceLevel 2 of 5Typical for packaged food brands of similar revenue and headcount: level 2 of 5.Core systems connect through APIs or simple automations, so an AI agent could run a workflow step once one is added.
- Team enablementHigh confidenceLevel 2 of 5Typical for packaged food brands of similar revenue and headcount: level 2 of 5.A company-approved AI tool with some trained users.
- GovernanceHigh confidenceLevel 3 of 5Typical for packaged food brands of similar revenue and headcount: level 1 of 5.Policy plus data classification plus a rule for human review of AI output.
The ladder · four stages
What AI can do here
- 01
Chat and Copilot
People use AI chat tools on their own work, one task at a time.
- 02
Connected Intelligence
AI answers questions from the company's own data.
- 03
Workflow AgentsYou are here
AI runs a step of a workflow, and a person reviews it.
- 04
Orchestrated Systems
Agents run work across functions, and people handle the exceptions.
What it takes Several steps across two functions run by agents with evals and cost per task tracked.
Ten use cases · effort to value
One to do first
Do first
- 01L3Retailer deduction matchingConnect
Plan
- 03L2Sales questions answered from the warehouse (needs foundation work first)Buy
- 04L3Production schedule drafting (needs foundation work first)Build
Quick wins
- 05L3Supplier invoice intakeBuy
- 06L3Broker email triageConnect
- 07L2Food safety record lookupBuy
- 08L2Retailer scorecard summariesConnect
- 09L2New hire answersBuy
Skip for now
- 10L3Freight invoice auditBuy
ML and planning, ranked with the rest
- 02MLReplenishment from retail sales data (needs foundation work first)Build
Ranked by value first, then effort, so a high-value use case that takes longer can sit above a quicker one. Value and effort are estimates; each use case's page gives the range and the assumption behind it. A ringed dot needs foundation work first. A square dot is a model, not an agent; it ranks with the rest and is listed last.
How this compares
Against packaged food brands of similar revenue and headcount, this company is ahead of what's usual on data foundation, context layer and governance. It isn't behind on any dimension, and it's in line on agentic readiness and team enablement.
Industry snapshot: Blue Orange's analysis of public information on packaged food brands of similar revenue and headcount, refreshed October 2026. Its sources are on the last page.
- The verdict in one lineWhat the company can start with, in which department, and who checks the work.
- Where the value isThe value a year as a range, labelled estimate, with the assumption one tap away.
- What to do firstOne move with an owner and its cost for the first 30 days.
Every source is linked and every figure is an estimate with its assumption beside it.
Read the whole sample reportWho it's for
 | Operating partners | Run it on one company in a board prep, or send the link to every company in the portfolio and read the results side by side. | Assess a company |
|---|---|---|---|
 | Company teams | A CEO, COO or head of technology answers for their own company. The firm sees the report if they sent the link; otherwise it's yours. | Answer for your company |
 | Before a deal closes | Run it on a target from public information alone, and nothing is sent to the company. The report says so on every page. | Assess a target |
For firms
Partners at a firm on Blueprint sign in with their firm email and see the full report for every company that has run it, with the verdict, the five levels and the top three use cases side by side. Send the assessment with the firm's name and logo on it, or run one yourself. We'll set a firm up in a day.
- Already on Blueprint?Sign in
- Not yet?Request free firm access
Questions
Is it free?
Yes. Access is by invitation: a firm sends you Blueprint, someone shares a report with you, or you request access and we let you in. Reading, sharing and the PDF cost nothing.
How long does it take?
About ten minutes to answer, then three to five minutes while the report writes itself. Keep the page open; it opens there.
What do you do with my responses?
They score the report you're reading. Anonymised, they feed an industry benchmark; nothing names the company. We don't sell the data and we don't email the company.
Who sees the report?
You, the people you share it with, and the firm that sent you Blueprint, which sees the full report.
How accurate is it?
The research is from public sources and every claim links to one, and you can correct the facts before the questions. AI writes the report, so treat it as recommendations, and ask Blue Orange to verify anything you'll act on.
Is it a sales pitch for Blue Orange?
No. Blue Orange built Blueprint, but the report's routes name real products and nothing in it requires us. It's a first read from public sources and your responses, so check it before you act. Checking it against the systems themselves is a free two-hour Diagnostic with a senior engineer, if you want one.
Does a target find out we ran it?
No. Research reads public sources only, and nothing is sent to the company. The report is yours, and the firm's if a firm sent the link.
Can we run it on a company we don't own yet?
Yes. Choose "Before a deal closes" on the start page and the report runs from public information only.