Back to blog
✍️Nolann Bougrainville

How to Prepare Your SME for AI Before Choosing Any Tool

You run an SME and people keep talking to you about AI and automation. Vendors promise spectacular time savings, but when it’s time to act, a simple question comes up: is your organisation really ready to make use of these tools?

In many SMEs, the problem is not a lack of solutions; it’s that the company isn’t structured to benefit from them. The result: projects drag on, teams get lost, and after a few months everyone goes back to the old way of working.

In this article, we’ll look at how to prepare your SME for AI before you even choose a tool. The goal is to build a solid foundation so that your future AI and automation projects are easy to launch, clear for your teams, and genuinely useful in daily operations.

1. Why so many AI projects fail before they start

People often blame the “wrong solution” or the “wrong provider”. In reality, many AI and automation projects fail for much more down‑to‑earth reasons:

  • Processes are unclear or live only “in people’s heads”.
  • Decision rules change depending on who is in charge or on the day.
  • Information is scattered across emails, files, and multiple tools.
  • Teams have neither the time nor the framework to test new ways of working.

AI does not invent your organisation. It amplifies what already exists. If your processes are fuzzy, it amplifies the fuzziness. If they are clear, it amplifies efficiency.

Preparing your SME for AI is therefore not about installing software. It is first about making your ways of working more explicit, shareable, and manageable.

We’ll do this through four very concrete workstreams, accessible even if you are not technical.

2. Clarify your “rules of the game” before talking technology

Before asking “which AI tool should we choose?”, ask instead: which rules do we want to automate or have AI assist with?

2.1. Make recurrent decisions explicit

Start by identifying 3 to 5 decisions that come up all the time, for example:

  • Approving or rejecting an exceptional discount for a client.
  • Prioritising incoming requests (support, internal projects, sales requests).
  • Approving holiday, training, or expense requests.

For each one, write down:

  • The criteria used (amount, type of client, urgency, history, etc.).
  • The implicit rules (what you often say: “below X we approve”, “contract customers are always prioritised”, and so on).
  • Frequent exceptions (sensitive cases, strategic clients, special context).

The goal isn’t to freeze everything but to make visible what already guides your choices. This is the raw material that will allow AI to help you prepare, filter, and prioritise later on.

2.2. Turn these rules into one‑page guides

For each recurrent decision, create a short one‑page guide:

  • Context: when is this rule used?
  • Steps: 3 to 5 steps, no more.
  • Who decides? Who executes?
  • What must remain human (for example: assessing a sensitive situation, validating a major commercial gesture).

This may seem basic, but it changes everything: you move from purely “gut‑feel” decisions to a clear framework that can be:

  • explained to your teams,
  • tested,
  • then partially automated or supported by AI.

3. Structure your information so AI can actually help

An AI system will not magically “guess” your emails, files, or unwritten rules sitting on a server. To be useful, it needs access to a minimum of organised information.

3.1. Bring together the information used for decisions

Choose a limited scope (e.g. holiday management, client reminders, internal requests) and answer three questions:

  1. Where is the information today? (emails, Excel, Google Drive, business software…)
  2. Who needs access to it? (owner, manager, assistant, whole team…)
  3. In what format would it be easier to consult? (one table, a standard request form, a shared folder…)

Then apply a simple rule:

One decision = one main source of truth.

For example:

  • All holiday requests go through a single form, no more scattered emails.
  • All late payments are visible in a single table, not ten different files.

This is still not AI, but without it, any AI tool will simply mirror your existing disorder.

3.2. Make your processes visible

A simple diagram is enough to clarify what actually happens. Here is a typical internal request flow (HR, IT, admin) before automation:

Rendering diagram...

After preparation:

  • Requests go through a single form.
  • Key pieces of information are standardised.
  • Follow‑up is visible to everyone.

This simple change makes it much easier to add AI later to: classify requests, suggest standard answers, detect urgent cases, and so on.

4. Build a small‑scale experimentation culture

One of the main barriers to AI in SMEs is not technical but cultural:

  • Fear of “breaking” existing workflows.
  • The feeling that everything must be a big project.
  • Lack of time to experiment.

Preparing your SME also means introducing a way of working that accepts small, low‑risk experiments.

4.1. Define a simple and reassuring test framework

For each future AI or automation experiment, decide upfront:

  • Limited scope: one team, one type of request, one process.
  • Test duration: 2 to 4 weeks, no more.
  • Simple indicators: time saved, errors avoided, team comfort.
  • Safety rule: humans always keep the final decision.

Explain to your teams that the goal is not to replace them, but to remove repetitive, error‑prone work. And that any test can be stopped without drama.

4.2. Appoint an “AI use‑cases lead”, not a technical expert

You do not need a data scientist to get started. What you do need is a person in charge of AI use cases (or a pair of people) whose job is to:

  • collect pain points and automation ideas,
  • prioritise topics with management,
  • coordinate experiments with your partners (internal or external),
  • gather feedback from the teams.

This person does not need to code. They must understand the business, listen, and structure needs. They are the bridge between your real‑world operations and the solutions offered by providers.

5. Build a minimum foundation before calling an AI provider

Once these basics are in place, you will approach partners with a much clearer view of what you expect.

Before launching a tender or project, make sure you have at least:

  • 1 to 3 processes described in simple terms (even in Word or on a whiteboard).
  • For each one, an identified main data source (file, tool, CRM, etc.).
  • A few decision rules written down (criteria, exceptions, sensitive cases).
  • A clear idea of what must remain 100% human (sensitive situations, key clients, high‑impact decisions).

With this foundation:

  • You will ask providers the right questions.
  • You’ll avoid oversized projects.
  • You’ll keep control over priorities and the value created.

Preparing your SME for AI is not about “waiting until everything is perfect”. It is about putting in place a minimum of clarity and structure so that each future project is simpler, faster, and more useful.

Practical section: 10‑day action plan to prepare your SME for AI

Here is a straightforward action plan you can apply as is, to build this foundation without spending months on it.

Days 1–2: choose your first scope

  • Quickly list the areas where AI or automation could help (HR, client reminders, support, internal admin, etc.).
  • Choose one single scope to start with, for example: internal request management, overdue invoice follow‑up, or holiday approvals.

Days 3–4: map the current process

  • Describe the process in 5 to 7 steps at most.
  • Identify who does what, with which tools, and using what information.
  • Note down the pain points: delays, duplicate work, errors, tensions.

Days 5–6: make decision rules explicit

  • For this process, which decisions are made regularly?
  • What criteria are used (amounts, deadlines, status, type of client, etc.)?
  • What frequent exceptions require human attention?

Put all this into a simple document shared with the people involved.

Days 7–8: simplify and centralise information

  • Decide where the main source of truth will be (a file, a shared tool, a form).
  • Reduce the number of channels used: ideally, a single entry point (one form per request type, one table for follow‑up, etc.).
  • Implement this minimum setup without overhauling your whole IT stack.

Days 9–10: define your AI experimentation framework

  • Appoint an AI use‑cases lead (or a pair) for this scope.
  • Set your test indicators: time saved, errors avoided, team satisfaction.
  • Draft a one‑page preparation brief including:
    • Chosen scope
    • Simplified process
    • Decision rules
    • Available data
    • What must remain human

This brief will become the basis for discussions with any future provider or partner.

Conclusion

To sum up, before choosing any AI or automation tool, you can already:

  • Clarify your rules of the game for recurrent decisions.
  • Structure information around one source of truth per process.
  • Introduce a culture of small, low‑risk experiments your teams can trust.
  • Appoint an AI use‑cases lead to bridge business needs and solutions.
  • Prepare a clear initial brief that partners can act on.

These actions require neither technical skills nor heavy investment. Yet they make a major difference between AI projects you suffer through and AI projects that genuinely help your business.

If you would like support in your digital transformation, Lyten Agency can help you identify and automate your key processes. Contact us for a free audit.

Back to blog