Back to blog
✍️Xavier Vincent

Turn your internal procedures into a simple AI-powered operating system for your SME

You run an SME and your ways of working haven’t really changed in years… while everyone keeps talking about AI and automation. You keep hearing that “it will revolutionise your business”, but in practice you don’t see how to start without a big IT project. This article offers a very simple entry point: using AI to standardise your internal procedures and turn them into a real operating system for your company.

We’ll look at how to transform your operating procedures, scattered checklists and “things people keep in their heads” into a living, AI-assisted system. The goal is not to automate everything, but to make your ways of working clear, shared and easy to execute for every team. By the end, you’ll have a step-by-step method and a practical checklist to launch a first scope in a few weeks, without jargon or complexity.

Why your current procedures are not enough

In many SMEs, procedures do exist… but they don’t really drive day-to-day work. Typical situations include:

  • Word or PDF documents that nobody opens
  • Excel checklists hidden in shared folders
  • “Unofficial procedures” known only by a few key people
  • Variants for specific clients or exceptions that are never documented

The consequences are familiar:

  • New hires take a long time to become autonomous
  • Quality varies from one person to another
  • The same mistakes keep coming back
  • Managers are constantly asked to answer questions they’ve already settled before

A procedure that is not used in daily work is not a procedure – it’s an archive.

AI changes the game not by magically writing procedures for you, but by turning those procedures into a concrete assistant: a system that helps your teams apply the right steps, at the right time, with the right checks.

From procedure to “operating system” for your SME

The key idea is simple: your procedures should no longer be just text. They should become the backbone of an AI-assisted way of working.

In practice, this means:

  • Starting from your existing procedures (even if they’re imperfect)
  • Clarifying and structuring them
  • Letting AI prepare, personalise and remind the key steps
  • Automating the small repetitive tasks around these procedures

You move from:

  • A static document → to an operational assistant that guides the user
  • Dependence on a few experts → to shared and secured know-how
  • Repeated mistakes → to systematic checks built into the workflow

Concrete example: onboarding a new employee

Take a very common process: a new hire joining the company.

Today, it often looks like this:

  • A more or less up-to-date HR checklist
  • Last-minute emails sent to IT, the manager, reception…
  • Regular oversights: badge, software access, equipment, onboarding plan…

With an AI-assisted “operating system” approach, you get:

  1. A clear procedure broken down into steps (before arrival, day 1, first week, first month)
  2. A simple form where you enter the key information about the new hire
  3. An AI assistant that automatically generates:
    • A personalised checklist for each department
    • Standard emails (to be reviewed) to notify the right people
    • A generic onboarding plan to be adapted to the role
  4. Automated reminders for critical tasks (contract, trial period, access to sensitive tools…)

Your teams still take the decisions, but the mechanics are handled by the system. AI becomes the assistant that prepares, reminds and structures the work.

Rendering diagram...

This diagram shows how an existing procedure becomes the heart of a tooled workflow: you keep the business logic, AI orchestrates the execution.

How to structure procedures so AI can actually help

AI works very well with clear rules and explicit steps. The first move is therefore not technical but methodological.

1. Pick a single business process

Resist the urge to redesign everything. To start, pick a process that is:

  • Frequent (at least several times per month)
  • With stable, repeatable steps
  • That creates frustration when it goes wrong (errors, delays, tension)

Examples:

  • Employee onboarding
  • Creating a new client or supplier
  • Kicking off a new project
  • Handling a standard customer request (support ticket, contract change, etc.)

2. Describe the process as if you trained a newcomer

Forget about AI for a moment. Ask yourself: “If I had 30 minutes to train someone on this process, what would I tell them?”

You want to capture:

  • The main steps (3 to 7 at most)
  • For each step:
    • Objective
    • Required information
    • Actions to perform
    • Checkpoints (what must not be forgotten)

A simple structure, in Word, Markdown or your usual tool:

  • Step 1: …
    • Objective: …
    • Inputs: …
    • Actions: …
    • Checks: …

3. Identify what can be prepared, generated, reminded

For each step, ask three simple questions:

  • Prepared by AI? (e.g. gathering information, summarising an email, extracting data from a document)
  • Generated by AI? (e.g. email draft, contextualised checklist, report template)
  • Reminded automatically? (e.g. follow-up after 7 days, check before a deadline, manager approval)

Very quickly, you’ll see “assistance zones” where AI can take work off your teams’ plate without replacing their judgement.

Implementing a first scope in 10 steps

You don’t need to build your own software. Using existing tools (AI chat, online forms, lightweight automation), you can build a first “mini operating system” in a few weeks.

Here is a framework we often use at Lyten Agency with our SME clients.

Step 1 – Choose the pilot process

  • Bring together 2–3 people who actually run this process
  • Agree on the most frustrating but under control process (not an exceptional case)

Step 2 – Map the current process

  • Sketch the steps on a sheet of paper or a whiteboard
  • Note the pain points: lost information, retyping, frequent errors

Step 3 – Write a simple procedure

  • Formalise the steps as described above
  • Keep it to 2 or 3 pages: the goal is usage, not perfection

Step 4 – Create a single intake form

  • Create a simple form (Google Forms, Typeform, your internal tool) that gathers the information needed to start the process
  • Goal: stop chasing missing information

Step 5 – Configure an AI business assistant

  • Using the procedure, configure an AI assistant (within your chat tool or a secure environment) with:
    • Business context
    • Step-by-step procedure
    • Examples of good outputs (typical emails, checklists, summaries)
  • The assistant should be able to:
    • Suggest the right steps for each case
    • Draft messages
    • Produce contextualised checklists

Step 6 – Automate the first micro-tasks

Without going too far, you can already:

  • Automatically create a task in your project tool (Trello, Asana, Notion…)
  • Send a standard email based on the form information
  • Generate a structured folder in your drive with the right templates

Step 7 – Test on 3 to 5 real cases

  • Ask teams to use the system on real work items
  • Note what is helpful and what is painful
  • Collect feedback: clarity, time saved, errors avoided

Step 8 – Adjust the procedure and the assistant

  • Refine the wording
  • Add concrete examples
  • Remove unnecessary steps

Step 9 – Set rules for usage

  • When must the system be used? (always, above a certain amount, for specific client types…)
  • Who owns procedure updates?
  • How are problems and exceptions reported?

Step 10 – Measure and decide on expansion

  • Over 1–2 months, measure:
    • Average time per case
    • Number of errors/complaints
    • Perceived clarity for the teams
  • Decide whether to:
    • Extend to similar processes
    • Go deeper on automation for this one

Practical section: checklist to launch your AI-assisted operating system

Use this checklist as a quick guide before you start.

Kick-off checklist

  • [ ] I have chosen one single pilot process, frequent and well understood
  • [ ] I have brought together 2–3 key users of this process
  • [ ] We have described between 3 and 7 steps
  • [ ] Each step has: an objective, inputs, actions, checks
  • [ ] We have identified at least 3 tasks that can be prepared or generated by AI
  • [ ] A single intake form has been defined
  • [ ] An AI business assistant has been configured using this procedure
  • [ ] We have set up at least one simple automation (task, email, folder)
  • [ ] We plan a test on 3 to 5 real cases
  • [ ] We have defined simple indicators (time, errors, team feedback)

What you can do this week

  1. Book a 60-minute session with two colleagues to pick the pilot process and sketch its main steps.
  2. Draft a first version of the procedure in a shared document.
  3. Configure a first (even basic) AI assistant with this procedure to test how it can:
    • Suggest steps
    • Generate a checklist
    • Prepare a standard email
  4. Identify one basic automation (task creation, automatic email) you can set up with your existing tools.

By progressing in small, concrete steps, you gradually build a real operating system for your SME, rooted in your actual ways of working and amplified by AI.

Conclusion

By structuring your procedures so AI can use them, you’re not launching “another AI project”. You are equipping the way you work. Your teams gain clarity, errors decrease, onboarding becomes smoother, and you reduce dependency on a few key people.

Keep three core ideas in mind:

  • AI is most useful to prepare, generate and remind around clear procedures
  • The real win is not the tool itself, but the creation of a simple operating system for execution in your SME
  • One well-chosen pilot process is enough to prove the value and bring people on board

If you’d like support with your digital transformation, Lyten Agency helps you identify and automate your key processes. Contact us for a free audit.

Back to blog