Most SMEs that come to me saying “we want to use AI” do not have a technology problem. They have a decision problem. They have read the articles, attended the webinars, collected a list of possibilities, and now they are stuck. Not because the options are bad, but because nobody has helped them work out which ones to pursue first, whether their internal foundations can support them, or what governance should look like before they start.
This case study covers an engagement with a Mauritian professional services firm that found itself in exactly that position. The firm was growing, operationally stretched, and genuinely interested in AI. What they lacked was not ambition. It was structure.
A growing firm with the right instincts and no clear framework
The client is a Mauritian SME operating in professional services, with a team of 30 to 50 people across operations, client delivery, and administration. They had been in business for over a decade and had built a solid reputation in their sector.
Operationally, they were feeling the weight of growth. Manual processes that worked at a smaller scale were becoming bottlenecks. Key staff were stretched across too many responsibilities. There was increasing pressure to do more with the same headcount.
Several team members had started exploring AI tools independently. A few had experimented with chatbots for internal queries, others with document drafting tools. The managing director had attended two industry events that mentioned AI prominently but came away with more questions than answers.
What they did not have was any structured view of what AI could do for their specific operations, which areas were ready for it, or what they should try first.
Scattered ideas, unclear priorities, and no governance
The core issues were typical of SMEs at this stage. I see them repeatedly across different sectors in Mauritius.
Fragmented priorities
Different team members had different ideas about where AI would help most. Finance wanted automation. Client services wanted better response times. Operations wanted workflow optimisation. Nobody had compared these against actual readiness or business impact.
Weak process clarity
Many of the firm's internal workflows were undocumented or carried entirely in people's heads. That matters because AI works best when processes are defined and repeatable. If you cannot describe the workflow, you cannot automate it.
Uneven data foundations
Some departments had clean digital records. Others still relied on spreadsheets, local files, and manual tracking. The data was scattered across systems with no central structure. This made it hard to know where AI could be applied without significant preparatory work.
No governance thinking
Nobody had asked the basic questions: Who decides which AI tools get adopted? What data can be fed into third-party models? What happens when a tool generates incorrect output? These are not optional considerations. They are prerequisites.
Structured advisory, not a technology pitch
I ran this engagement over four weeks, working directly with the managing director and a small group of departmental leads. The work followed a structured sequence that I have refined through several similar engagements.
Readiness assessment
I mapped the firm's current state across four dimensions: technology infrastructure, organisational culture, process maturity, and data quality. This is not a tick-box exercise. It involves structured conversations with staff, a review of existing systems, and an honest appraisal of what the firm can actually support right now, not what it aspires to support in eighteen months.
Use case identification and prioritisation
I collected every AI idea the team had considered and added several they had not. Then I scored each one against feasibility, business impact, data readiness, and implementation complexity. The goal was to move from a long list of possibilities to a short list of realistic starting points.
Gap analysis
For each prioritised use case, I identified the specific gaps that needed closing before implementation could begin. Some gaps were technical. Others were about process definition, data cleanup, or staff capacity. This step prevents the common mistake of selecting a use case and then discovering halfway through that the foundation is not ready.
Governance considerations
I drafted a practical governance brief covering data handling, vendor evaluation criteria, accountability for AI-generated outputs, and escalation protocols. This was not a 40-page policy document. It was a working framework appropriate for a 40-person firm.
Phased roadmap
I built a sequenced roadmap with three phases, each containing specific actions, dependencies, owners, and success criteria. The roadmap accounted for the firm's actual budget, team capacity, and risk appetite. It was designed to be executable, not aspirational.
Tangible outputs, not slide decks
By the end of the engagement, the firm had a set of working documents they could act on immediately:
AI readiness scorecard
A structured assessment across four dimensions with specific scores, commentary, and recommended improvements for each area.
Prioritised use case matrix
Eight use cases evaluated and ranked. Three were recommended for immediate exploration, two for medium-term development, and three were flagged as premature given current readiness.
Readiness gap report
Specific gaps per use case, with estimated effort and cost to close each one. This helped the firm budget realistically rather than guessing.
Governance considerations brief
Practical guidelines for data handling, vendor selection, output accountability, and escalation. Sized for an SME, not a multinational.
Phased adoption roadmap
Three phases over 12 months, with dependencies mapped, owners assigned, and clear criteria for moving from one phase to the next.
Clarity replaced noise
I am careful about outcome claims. This engagement did not produce a 300% ROI or transform the business overnight. What it did produce was something more practical and, I would argue, more useful.
The leadership team moved from competing, unranked AI ideas to a shared, prioritised view of what to pursue and in what order.
Two use cases that had seemed attractive on the surface were deprioritised after the readiness assessment revealed foundation gaps that would have been expensive to discover mid-project.
The firm began the first phase of the roadmap within three weeks of receiving the final deliverables, starting with process documentation and data cleanup rather than tool procurement.
The governance brief gave the managing director a framework for evaluating AI vendor proposals rather than making decisions based on demonstrations and sales pitches.
Internal AI conversations shifted from “we should try this tool” to “does this fit our roadmap and are we ready for it?”
Whether these outcomes lead to measurable financial returns will depend on execution over the coming months. I do not claim to know that yet. What I can say is that the firm is now making AI decisions from an informed position rather than an excited one.
Small economies need different advice
Most AI adoption guidance is written for large organisations in large markets. It assumes dedicated innovation teams, substantial budgets, and a tolerance for failed experiments that SMEs in Mauritius do not have.
When your team is 40 people and your operating margin is tight, you cannot afford to spend six months on a pilot that teaches you nothing except that your data was not ready. You need someone to tell you that before you start spending.
The firms I work with in Mauritius share common characteristics: they are capable, operationally competent, and cautiously interested in AI. What they typically lack is a structured way to evaluate their own readiness and a method for sequencing adoption that accounts for their actual constraints.
That is the gap I fill. Not as a software vendor or an implementation partner, but as a strategy advisor who helps you make better decisions about if, where, and when to adopt AI.