AI Readiness for Small Businesses: What to Put in Place Before AI’s First Task

title banner: AI Readiness for Small Businesses [Video]

AI readiness starts before anyone on your team opens ChatGPT, Copilot, or any other AI tool.

Here’s what usually happens instead. A business buys a few licenses, throws some tasks at them, and decides how it feels about AI based on whatever comes back.

Think about how differently you’d treat a new hire. You’d plan their first week, show them how the business works, give them access to tools, and only then hand them small tasks.

I teamed up with JD Jankowski of CelereTech, a managed IT provider that serves as the IT department for small businesses, for The Blueprint for AI Organizational Readiness.

JD covered the technical side: keeping your data safe and deciding which tools your team can use. I covered the other half: picking the right first task and giving AI the context it needs to do good work.

The full replay is below. Under it, you’ll find the key takeaways and an AI starter list you can bring to your next team meeting.

How to Choose Your First AI Task

JD opened the roundtable with a common question: where do you start?

For months I’ve used three main rules. Lately I’ve added two more.

A good first task for AI:

  1. Happens often. That’s where the time savings add up.
  2. Has clear instructions. They can be written down or just in your head, as long as you could explain them.
  3. Is something you or your team would rather not do. People are motivated to hand off work they already dislike.
  4. Connects easily to AI. If the task lives in a tool AI can reach, setup gets much simpler. I’ve started choosing tools partly for this reason.
  5. Is quick to review. If checking AI’s work takes as long as doing it yourself, pick a different task.

JD added one more from the IT side: choose something that’s easy to undo. If the pilot doesn’t work out, you want a clean exit.

Give every AI task an owner who knows the subject. AI delivers answers with total confidence, even when they’re wrong. A reviewer who doesn’t know the topic well will read it and think, “Sounds great.” An expert catches it when AI starts drifting off course. That’s why the owner should be the expert.

JD’s addition: that owner also decides what data the tool can see and shuts off access when someone leaves the company.

How do you know if AI is working? Decide what success looks like before you start. Most AI goals fall into one of three buckets:

  • Save time
  • Get better output
  • Do something you couldn’t do before with your current team

If the goal is time, decide how you’ll measure it. Expect the first few rounds to take longer while you fine-tune the process.

If the goal is better output, write down what “better” means. Maybe it sounds like your brand, gets the facts right, and speaks to your customers. Whatever it is, decide it up front so you have something to measure against.

Business AI Readiness Starts With the Foundation

JD’s section came down to one idea: AI is only as strong as what you build under it. He started with four basics.

  • Reliable tech. Most AI tools run in the cloud, so you need dependable internet and computers that are updated and secure.
  • Strong passwords and multi-factor authentication (MFA). MFA is the second check when you log in, like a text code or an authenticator app. Microsoft research found it cuts the risk of an account being compromised by more than 99%.
  • Trained people. Your team needs to know how to spot AI-powered scams and how to use the tools you give them.
  • Organized data. Garbage in, garbage out. AI’s answers are only as good as the information you give it.

Does AI use your company data for training? It depends on the tool and the plan. ChatGPT’s personal plans (Free, Plus and Pro) can use your conversations for training unless you turn that off, while OpenAI says its business plans don’t by default. Microsoft, Anthropic and other vendors set their own rules, so ask each one.

Either way, anything you type, paste or upload now lives on the vendor’s servers. How long they keep it ranges from a year to five years, depending on the company and your agreement. Ask which of their partners can see it, too.

Is it safe to use AI for business? It’s safest on a business account. Business plans typically keep your data out of model training, let you set how long data is kept, and give you admin control. When an employee leaves, you can switch off their access. A personal account gives you none of those controls.

What information should you keep out of AI? Anything confidential. JD’s tip is to sort your data like a traffic light:

ColorWhat It IncludesAI Rule
Red: confidentialCustomer records, health data, payment card numbers, passwords, HR filesKeep out of AI tools
Yellow: internalProcedures, proposals, quotes, client listsApproved business tools only
Green: publicWebsite content, marketing materials, job postingsFine for any approved tool

Remember, industry rules like HIPAA (health information) and PCI (card payments) still apply. So do laws like Illinois’ Biometric Information Privacy Act, which can cover AI tools that use someone’s voice. If your client contracts or NDAs limit sharing with third parties, an AI vendor counts as a third party. Check your cyber insurance policy for AI exclusions too.

Five Questions to Ask Any AI Vendor

  1. Is my data used to train your AI?
  2. How long do you keep it, and can a contract change that?
  3. Where is it stored, and who can access it?
  4. Can we delete it when we leave?
  5. Will you put it in writing, such as a business associate agreement (BAA), and share an independent security audit like a SOC 2 report?

Vendors change their terms often, so ask again even if you checked a year ago.

Give your team an AI menu. Your team is probably using AI already. According to Microsoft and LinkedIn’s 2024 Work Trend Index, 78% of AI users bring their own AI tools to work. Unapproved tools the company can’t see are called shadow AI.

JD’s fix is a simple menu:

  • Approved: business accounts only. Yellow data is OK.
  • Approved with limits: low-risk uses like brainstorming, with green data only.
  • Never allowed: company data in personal accounts, and browser add-ons that can read everything on your screen.

Every tool gets a named owner. Give your team an easy way to request new tools, and review the menu every quarter.

Remember that AI sees what you see. An AI tool inherits the permissions of the person using it.

JD’s example: if a payroll folder was accidentally left open years ago, an AI tool connected to that employee’s account can see payroll too. Give people only the access they need, and ask two questions about every shared folder: who can see it, and why?

Do you need an AI policy for employees? Yes, and one page is enough. It covers six things:

  1. Your AI menu of approved tools
  2. Rules for which data goes in which tool
  3. Work accounts only
  4. A person checks AI’s work before anything goes out
  5. No new tool connections without approval
  6. An easy, no-blame way to report a mistake, because the sooner you know, the faster you can fix it

CelereTech is giving away a free AI acceptable use policy template, both blank and filled in, at celeretech.com/blueprint.

If you haven’t taken their AI Implementation Readiness Assessment yet, it’s a quick way to see where you stand: take the assessment.

Onboard AI Like a New Employee

Think of AI as the most knowledgeable, fastest-working employee you’ve ever hired. It knows a little about almost everything. The one thing it knows nothing about is your business.

With a new hire, the prep starts long before day one: training, job shadowing, access to the right tools, a checklist of what has to happen before they do real work.

AI needs the same setup. In my half of the webinar I focused on two parts of it: the right task size and the right context.

Start with one leaf. I borrowed this from Nate Herk’s book Becoming AI Native. Picture your whole job as a tree.

  • Limbs are the big parts of your job, like managing people, writing reports, running ad campaigns or producing a newsletter.
  • Branches are the tasks inside each limb. For a newsletter, that’s content research, planning, drafting, a brand voice check, design, and follow-up.
  • Leaves are the smallest steps. Newsletter research breaks down into scanning for new stories, filtering them for your audience, summarizing each one, and rating which ones to use.

Hand AI one leaf, or a small cluster of leaves. Just like a new hire, it earns bigger jobs by doing small ones well, and you build trust along the way.

How do you train AI on your business? You give it context, the same way you’d train a new hire. T.R.A.I.N.E.R. is the framework I use before handing AI any task.

When I say “train” in this post, I mean giving AI the right context. (Your information stays out of the AI model itself.)

Here’s how it works, using my own PinPoints newsletter research as the example.

LetterWhat It MeansNewsletter Research Example
TaskWhat you’re doing and whyFind recent industry news that matters to my readers. AI does better work when it understands the purpose.
RecipientWho the output is forProspects who know OnPoint Results but haven’t hired me yet
AssetsThe information, files and tools AI needsMy customer persona, business goals, and access to the sources I want searched
InstructionsHow to do the jobHow far back to look, checking past issues for repeats, weighing each story against my persona
NormsWhat good looks likePast issues, or newsletters I admire if I’m just starting out
Expected outputExactly what to deliverA short summary of each article, a few sentences on why it fits, and a link
ReviewA person checks the workI read the articles myself before choosing what goes in

Assets is the biggest piece, and it’s where most businesses come up short. Part three of the readiness assessment asked about exactly these: a documented customer persona, a positioning statement, your brand voice, your offers, and access to your reporting. AI can only use them once they’re written down.

Let AI write its own instructions. Already have written instructions for a task? Give them to AI and ask it to rewrite them for itself. It’s the one following them, so it writes them in a way it can use.

No instructions yet? Ask AI to interview you about how you do the task, then write them up. When you find yourself correcting the same thing twice, tell AI to update its instructions.

Show examples. AI follows a good example much better than a list of things to avoid. Give it two or three samples of finished work you’re happy with.

Watch how long review takes. Review is where a person catches problems and improves the process. Over time, your corrections should shrink. If review keeps taking as long as doing the work yourself, the task is too big or isn’t a good fit for AI yet.

Your next step: skills. A skill is a packaged set of instructions an AI tool can reuse. Anthropic, the company behind Claude, developed the format and released it as an open standard. It’s been adopted by a growing number of AI tools, including ChatGPT and Microsoft Copilot. A skill is a natural home for everything you gathered with T.R.A.I.N.E.R.

In the replay at 48:40, you can watch mine in action. I type one sentence, “Let’s work on the next PinPoints newsletter,” and my research skill returns a ranked list of stories, each with a summary, a link, and a note on why it fits my readers. Then I do the human review.

Your AI Starter List

Run through this starter list with your team to see where you stand. “In progress” is a fine answer, as long as you have a next step. When you want the detailed, scored version, take the full AI Implementation Readiness Assessment.

The Foundation

  • Computers are up to date, and MFA is on for every account
  • Company data is sorted into red, yellow and green
  • AI runs on business accounts, and company data stays out of personal accounts
  • You have an AI menu of approved tools, each with a named owner
  • Your AI vendors have answered the five questions, in writing
  • A one-page AI policy is written and shared with your team

The First Task

  • You chose it using the five criteria: frequent, clear, unloved, easy to connect, quick to review
  • An expert owner is assigned to review the output
  • Success is defined: time saved, better output, or something new
  • Your context is gathered with T.R.A.I.N.E.R.: persona, brand voice, offers, and good examples
  • You’re tracking how long review takes

Free AI Readiness Resources

Everything from the session is free, with no form to fill out:

You’ll find the scorecard and handout together at onpointresults.com/blueprint.

Start With One Task

AI readiness is a handful of practical decisions you can start this week. Pick one task. Sort your data. Write down what good looks like. Then hand AI a single leaf and see how it does.

If you’ve found a task that scores well and want a second set of eyes on how to set it up, let’s talk it through. Discovery calls are free, and you’ll be talking with me directly.

Book a Discovery Call

Nikki Stine, founder of OnPoint Results

About the Author

Nikki Stine is the founder of OnPoint Results, an independent marketing consultancy for small businesses. She started in 2018 helping local businesses show up in Google searches, and that work grew into full marketing partnerships. Today she helps owners make informed decisions and avoid random acts of marketing.

Since 2024, Nikki has tested AI in her own business first, then put it to work for clients. She gives presentations for business owners who want to get more out of AI, and she helps them find a clear next step and take it.