Before you buy the platform: a systems modernisation playbook for growing organisations

Growing organisations reach a predictable moment. The business has outgrown its spreadsheets, messaging groups and personal workarounds. Orders, stock, approvals and customer commitments are scattered across tools and people. Somebody proposes buying an ERP, a CRM or a "platform", and a vendor demonstration is scheduled.

This is the moment that decides whether the next three years are spent running a better business or maintaining an expensive one. The outcome rarely depends on which product is chosen. It depends on what was understood before the choice.

Fragmentation is a symptom

When information lives in many places, the cause is usually that the work was never designed as one flow. A purchase request begins in a chat, becomes a spreadsheet line, is approved verbally and reappears as an invoice that nobody can match to a delivery. Each step is reasonable. Together they have no owner and no record.

Replacing the tools without redesigning the flow moves the fragmentation into a new system. We have seen, and the research supports, the same lesson repeatedly: Deloitte's 2026 study found over 80% of senior IT executives confident in their ability to deploy technology at scale, while 75% expected their operating models and processes to need to change within 12 to 18 months. The technology is not the hard part.

A six-step sequence

1. Define the value in operating terms

State what should be different in numbers the business already cares about: days to close the month, time from enquiry to quote, stock accuracy, approvals completed on time. If the benefit cannot be stated plainly, the system is not yet justified.

2. Map the flow from first touch to final record

Follow one real transaction through the organisation, with the people who handle it. Mark each hand-off, each wait, each re-keying of data. The map usually reveals that a few hand-offs account for most of the delay and error.

3. Fix ownership and approvals before configuration

Systems enforce rules, so the rules must exist. Who may approve what, up to which value, and who delegates when they are away? Decisions that are vague on paper become arguments in a workflow engine.

4. Capture data once, at the point of work

Every duplicate entry is a future inconsistency. The best systems record events where they happen, on a phone in a workshop, at a receiving dock or on a quote form, and let everything else derive from that record, with an audit trail.

5. Build or buy to fit the flow, not the reverse

Off-the-shelf platforms are the right answer when the process is standard and the vendor's design is good enough. Custom components are justified where the process is a source of advantage. Many organisations need both: a proven core with carefully chosen extensions. The decision should follow the map, and be revisited as the business changes.

6. Roll out by role, and train as part of the build

Adoption is a design problem. Introduce the system role by role, make the first release valuable to the people who must change their habits, and train in the work, not in a classroom. A system that is correct but unused delivers nothing.

Where AI fits in a modernisation

Gartner expects a third of enterprise software applications to include agentic AI by 2028, up from under 1% in 2024. In practice this means that the platforms you adopt will increasingly arrive with AI inside. Two consequences follow. First, clean, consistent, well-owned data becomes more valuable, because AI features perform in proportion to the quality of the record. Second, governance matters earlier: who may the assistant act for, and who reviews what it decides.

The Rwandan context

Rwanda has made AI a stated national priority. Cabinet approved the National AI Policy in April 2023. In June 2026 it approved a National Artificial Intelligence Agency, Rwanda's first institution dedicated entirely to AI, with a mandate covering development, adoption, investment and governance, including in public administration, healthcare, education and agriculture. The Rwanda AI Scaling Hub, backed by international partners, has secured about 25 billion Rwandan francs, nearly $17 million, to promote adoption across the economy and the public sector.

ICT and Innovation Minister Paula Ingabire has said that around 70% of the policy is focused on skills. That emphasis is worth noting for any organisation modernising its systems: the binding constraint, as in other markets, is people able to specify, run and govern the technology. UNESCO-cited challenges include shortages of specialised talent and limited training data. Organisations that invest early in internal capability will be better placed than those that outsource the understanding along with the build.

A short checklist before any purchase

• Can we describe the target process on one page?
• Does each approval have a rule and an owner?
• Will data be captured once, where the work happens?
• Do we know which parts are standard and which are our advantage?
• Is there a plan to train people inside the work?
• Have we named who will run the system after the project ends?