Few subjects attract as much attention in business conversations as artificial intelligence, and few produce as much confusion. Some organisations are told that AI will transform everything. Others worry that it is risky, expensive or irrelevant to them. The truth for most growing businesses sits in between: AI is useful for certain well-defined tasks, unsuitable for others, and worth adopting only where it creates measurable value under proper oversight.
Start with the task, not the technology
The right question is not “how can we use AI?” It is “which tasks take too much time, produce too many errors or delay our customers, and could any tool, including AI, help?” Often the answer is a better process or a conventional workflow. Sometimes it is AI.
Where AI tends to work well
Sorting and classifying incoming enquiries. Emails, web enquiries and support requests can be categorised by topic, urgency or department and routed accordingly. A person still handles the enquiry. AI simply gets it to the right person faster.
Extracting information from documents. Invoices, delivery notes, application forms and similar documents can be read and key fields extracted for a person to verify. This reduces manual data entry, which is slow and error-prone.
Preparing first drafts for review. Routine reports, summaries of long records or standard responses can be drafted by AI for a person to check, correct and approve. The time saving comes from editing rather than writing from scratch.
Assisting customer support teams. AI can suggest answers from an organisation’s own knowledge base, helping support staff respond faster and more consistently. The staff member decides what is sent.
Summarising internal records. Long case histories, meeting notes or account records can be summarised so that a person can get up to speed quickly.
Identifying anomalies in data. Unusual transactions, unexpected changes in performance or entries that do not fit a pattern can be flagged for investigation.
Supporting knowledge retrieval. Staff can find relevant policies, procedures or past work by asking questions in plain language, rather than searching through folders.
Where caution is needed
AI produces errors, and it can present them confidently. It should not make decisions that affect people’s rights, finances or safety without meaningful human review. It should not provide professional advice in place of a qualified person. It should not be given confidential client information or personal data without proper authorisation and safeguards, particularly where the AI service is operated by a third party outside the organisation.
Keep a person accountable
For every AI application, a named person or role should be responsible for the output. If AI drafts a report, someone reviews and signs off the report. If AI classifies enquiries, someone monitors whether the classification is accurate. Accountability does not transfer to software.
Protect data from the start
Before adopting any AI tool, understand where data goes, who can access it, whether it is used to train the provider’s models and how long it is retained. Kenya’s Data Protection Act, 2019 applies to personal data processed by AI systems just as it does to any other processing. Confidential client information should not enter external AI systems unless the organisation has the right to share it and appropriate safeguards are in place.
Measure whether it works
Define what success looks like before deployment. That might be time saved per enquiry, reduction in data entry errors or faster response times. Measure the baseline, deploy carefully, then measure again. If the result does not justify the cost and the risk, stop.
Begin with one well-chosen application
The most successful AI adoption in growing businesses usually begins with a single, clearly defined use: classifying support emails, extracting data from supplier invoices or drafting routine reports. Once that works and the organisation understands the oversight it requires, further applications become easier to evaluate.
AI is a tool. Used on the right problems, with the right oversight, it can save real time and improve consistency. Used indiscriminately, it adds cost and risk. The discipline lies in knowing the difference.