Business AI-Readiness Checklist: 15 Points

Most failed AI projects fail not because of technology but because the business wasn't ready for it: the data lived in the receptionist's notebook, the process existed only in the owner's head, and expectations were "let it just do everything itself." We've put together a 15-point checklist we use in audits before starting projects. Run through it in an hour and you'll know exactly where to start with AI in your company, and whether it's worth starting now.

How to score: 1 point for each item if the statement is true of you. The score interpretation is at the end of the article.

Block 1. Data (points 1-4)

AI works with what it's given. No data, no system, just a pretty demo.

1. Your key data lives in systems, not in people's heads

Prices, availability, delivery terms, schedule, answers to common questions, all of it is recorded somewhere: CRM, spreadsheets, a database, at worst structured documents. If half the company's knowledge exists in the format "you have to ask Olena," the first step of adopting AI in your business won't be about AI but about capturing knowledge.

2. The history of communication with clients is stored

Messenger conversations, emails, call recordings. This is gold: from 200-300 real dialogues we extract typical questions, objections, and customers' phrasings, and the bot speaks the language of your buyers rather than the language of a manual.

3. The data is more or less clean

There aren't five duplicates of one client in the CRM, products have descriptions, deal statuses match reality. "More or less" is the key phrase: no one has perfect data, but if the base hasn't been updated in a year, the AI will honestly answer with outdated prices.

4. You know which data can be shown externally

There's an understanding of what's public (prices, terms) and what isn't (cost price, clients' personal data). Without this, the retriever will one day quote your internal markup to a client.

Block 2. Processes (points 5-8)

5. The process you're automating is repeatable

AI pays off on volume. If requests/leads/documents come in by the dozens per day, automation gives a multiplier. If a task arises twice a month, it's cheaper to do it by hand.

6. The process can be described on one page

Try it right now: "the client writes -> the manager clarifies X -> checks Y in the system -> answers with template Z or escalates." If the description comes out, the process can be automated. If at every step it's "well, depends" with no criteria, formalize it first.

7. You know the bottleneck

AI readiness is when you can say: "managers spend 3 hours a day on identical questions" or "requests at night wait for an answer until morning and 30% of them go cold." A concrete bottleneck = a concrete success metric.

8. There's a fallback to a human

It's defined who picks up the dialogue when AI can't cope, and how fast. Systems without escalation lose exactly the clients everything was built for.

Block 3. The team (points 9-11)

9. There's a project owner on your side

One person who answers the contractor's questions, makes decisions, and has 3-5 hours a week for the project. Without them the project stretches out twice as long, proven.

10. The team doesn't sabotage

Managers understand that AI takes away the routine, not their salary. A practical test: ask the team which tasks they'd happily hand off to a robot. If there's a list, you have allies. If it's "everything we do is unique, nothing can be handed off," start with internal communication.

11. Someone can update the knowledge base

Prices change, the assortment updates, promotions end. You need a person who refreshes the information once every week or two, that's 20-30 minutes, but without it the bot degrades within a month.

Block 4. Money and expectations (points 12-15)

12. There's a budget for a pilot, not only for "everything at once"

A realistic entry into custom AI: $600-1,000 for a parser or simple automation, from $1,000 for a chatbot, from $3,000 for a CRM with AI modules. Plus $40-120/mo in operating costs. If the budget is $200, start with builders and ready-made tools, that's also a normal first step.

13. You calculate ROI rather than "we want AI"

A simple formula: (hours freed up x cost of an hour + previously lost leads x conversion x average check) versus (development + maintenance). If the math doesn't add up on paper, it won't add up in real life either. How our clients calculate ROI is in the case studies.

14. Expectations: 80-90%, not 100%

AI closes most of the flow and makes mistakes on the tails. If for you one wrong answer in a hundred is a catastrophe, you need an architecture with a human in the loop, and that's a different budget. But if right now clients simply get no answer at night, even 80% automation is a leap.

15. You're ready to give the system 2-4 weeks to settle in

The first weeks after launch are fine-tuning on real dialogues: scenarios get added, phrasings get fixed. Teams that budget for this period get a system that works for years. Teams that expect perfection on release day get disappointed by day three.

Scoring

Score	What it means	What to do
12-15	Ready. The bottleneck is known, the data is there	Launch a pilot on the most painful process
8-11	Partially ready	Start with something simple (a knowledge-base bot, a parser), close the gaps in parallel
4-7	Too early for custom	2-4 weeks to get the data and processes in order, then AI lands on prepared ground
0-3	AI won't help right now	First, basic systematization: CRM, knowledge capture, process descriptions

An important nuance: a low score isn't a verdict but a saving of money. A project launched at 5 points will cost you double: first development, then rework.

And one more thing about honest self-assessment. In audits we constantly see two distortions. The first is optimistic: "we have the data" means there's a spreadsheet somewhere last opened in March. The test is simple: try right now to find the answer to five typical client questions in two minutes, if you can't, neither will the AI. The second distortion is pessimistic: "everything we do is unique, nothing is structured" often means that 70% of requests boil down to a dozen scenarios anyway, no one just counted. A week of counting real requests (category, frequency, time to answer) is more useful than a month of pondering AI readiness.

Where to start if your score is enough

Don't start with the hardest thing. The best first projects are the ones with a flow, simple logic, and a measurable result:

• A knowledge-base bot closes common questions 24/7, the fastest payback (chatbots are the classic first step).
• Request automation AI qualifies the lead and enters it into the CRM, the manager gets a ready card.
• Parsing and monitoring collecting competitor prices or data from sources, from $600, works on its own.

For the full picture of what we build for companies, from bots to autonomous agents, see the AI for business page.

If you've run through the checklist and want a second opinion, come to a free 30-minute AI audit. We'll go through these 15 points together on your real processes, show where AI pays off in months and where it doesn't (and honestly say if it's too early for you). Book here: maxicolabs.com/contact.