Africa

After AI Infrastructure, Who Owns the Workflow?

Africa's AI infrastructure will create business value only when companies redesign the processes around it, define human oversight and assign clear ownership for outcomes.

By Towncrier Editorial Desk

July 17, 2026 · 7 min read

Photo by Tara Winstead on <a href="https://www.pexels.com/photo/robot-pointing-on-a-wall-8386440/" rel="nofollow">Pexels.com</a>

OPINION

By Victor David

The views expressed in this article are those of the author and do not necessarily reflect the editorial position of Towncrier Africa. Opinion submissions are published following editorial review for relevance, clarity and factual support, but responsibility for the analysis and conclusions rests with the author.

Compute, cloud and connectivity create capacity. Business value appears only when someone redesigns the process, sets the limits and stays accountable for the result.

Imagine a law firm in Lagos adding a chatbot, a real estate brokerage in Accra deploying an AI assistant and a logistics company in Nairobi automating customer replies. All three tools could work as advertised, yet staff might still be copying customer details into spreadsheets six months later.

That gap can stop a promising AI pilot from becoming a working business system.

Africa’s AI debate is right to focus on physical infrastructure. Cloud regions, reliable power, broadband, local data capacity and technical skills determine which organisations can use advanced systems, and on what terms. TownCrier Africa’s recent coverage of Google’s infrastructure investment made that case clearly.

Infrastructure answers the first question. Once it exists, another follows it into every boardroom: who turns that capacity into a working process?

A growing company does not benefit simply because it can access a model. It benefits when a customer gets a timely answer, a qualified lead reaches the right salesperson, an appointment is logged correctly or an operational exception is flagged before it becomes a crisis. That is the difference between access to AI and ownership of an AI-enabled workflow.

The digitalisation gap is really a usage gap

This pattern is already visible in the wider digital economy. The International Finance Corporation’s 2024 Digital Opportunities in African Businesses report found that access to digital tools is widespread but intensive use is not: 86 per cent of African firms had access to tools such as mobile phones and the internet, while fewer than one in three firms that had adopted digital technology used it intensively for business purposes.

The gap varies by function. Payments moved fastest, with 62 per cent of firms using advanced digital tools for that purpose. Adoption across most other business activities was lower.

A company can be digitally connected and still run on disconnected hand-offs. A customer inquiry arrives by phone, moves to WhatsApp, gets copied into a spreadsheet and then depends on one employee remembering to follow up. Adding an AI assistant to one step makes that step faster. It does not repair the chain.

World Bank research on technology adoption draws the same line. Reliable infrastructure lets a firm upgrade; it does not guarantee that it will. The research argues that the focus should move past access and examine how technology is used inside specific business functions.

AI raises the stakes because it can generate a plausible-looking output without completing an accountable business task.

A model’s output is not a business outcome

Take a service company that receives inquiries through calls, web forms and messaging apps. An AI tool can summarise a message or draft a polite reply. A working system has to recognise the event that starts the process, gather the minimum information needed, follow approved rules, update the correct record, know when to bring in a person and leave a trail of what happened.

Skip those connections and the business has gained another interface, not an operational workflow.

The person who owns that workflow does not need to be an AI engineer. In a smaller company it might be the founder, operations lead, sales manager or head of customer experience. Their job is to define the result, decide what the system may do, name the source of truth, set the points where human judgement stays in the loop and check whether performance is improving.

Technology teams and implementation partners can build and run the system. Accountability for the business process has to stay with the business that owns it.

Six questions before an AI pilot

African businesses do not need to start with a sweeping transformation programme. One bounded workflow will show whether a proposed system is useful, safe and worth the money.

1. What result are we trying to change?

Name the problem and measure the baseline first. It might be slow response time, missed calls, dropped follow-ups, repetitive data entry or inconsistent booking. “Use AI” is not an outcome.

2. What event starts the workflow?

The trigger must be observable: a new inquiry, a missed call, an overdue follow-up, a submitted document or a changed customer record. A clear trigger makes the process testable.

3. Which information can the system trust?

Name the source of truth and give the system only what the task needs. It should not decide from a stale spreadsheet when the live record sits in the CRM. It should not get broad access because broad access was the easy setting.

4. Which actions are authorised?

Spell out what the system may do and what it must never do. It might answer approved questions, collect intake details, update a record or offer open appointment slots. Pricing exceptions, refunds, legal commitments, sensitive advice and unusual cases stay with a person.

5. Where does a person step in?

Build the escalation route from day one. Low confidence, conflicting information, an angry customer, a failed tool or a request outside policy each needs a defined hand-off.

6. What record tells us whether the pilot continues?

Log the action, hand-off and result, then compare them with the baseline using a few measures that matter. If the workflow cannot show what happened, no one can audit it, improve it or decide whether it should scale.

Build for the operating conditions you have

There is no single African operating environment. Connectivity, language, regulation, payment infrastructure, customer behaviour and organisational capacity differ between countries, sectors and even neighbourhoods.

The International Telecommunication Union put internet use across its Africa region at 38 per cent in 2024: 57 per cent in urban areas and 23 per cent in rural ones. The same report priced an entry-level mobile broadband plan at 4.2 per cent of gross national income per capita, still above the UN Broadband Commission’s two per cent affordability target.

Those numbers should shape the design. A workflow built for constant desktop connectivity will fail where users rely on mobile data. Critical steps may need low-bandwidth interfaces, asynchronous handling or a fallback to voice, SMS or a person. A system serving several markets may need different language, payment and escalation rules rather than one continental template.

Data governance belongs in the design too. The African Union’s Data Policy Framework calls for a digital environment that protects rights while enabling data-driven economic activity. In practice, that means collecting less data, controlling access, documenting automated actions and following the rules of every jurisdiction the system touches.

The AU’s Continental AI Strategy makes the same argument at a larger scale. It places infrastructure and skills alongside data, governance, research, innovation and the integration of AI into economic sectors. Funding infrastructure does not automatically create responsible use.

The missing layer is operational ownership

Africa needs continued investment in compute, connectivity, power and digital skills. It also needs people who can turn those foundations into dependable, everyday work.

That implementation layer includes business operators who understand the process, technical teams who can connect the systems, sector specialists who know where automation should stop and leaders willing to stay accountable once the demo is over.

The strongest AI projects on the continent will not necessarily have the flashiest interface or the most autonomy. They will finish a bounded task, keep human judgement where it belongs and produce a result the organisation can check.

Which brings us back to the imagined law firm in Lagos, brokerage in Accra and logistics company in Nairobi. None needs a bigger model. Each needs someone inside the business who owns the workflow from end to end and is still there to answer for it six months later.

Compute makes AI possible. Workflow ownership makes it useful.

Disclosure

The author is the founder of West Work Studios, an AI automation agency. This affiliation is disclosed in the interest of transparency.

Author biography

Victor David is the founder of West Work Studios, an AI automation agency that builds AI agents, voice systems, sales automation and connected business workflows for growing companies. His background in systems administration informs his focus on practical implementation, infrastructure and human-supervised automation.

Editor’s Note

This opinion article builds on Towncrier Africa’s reporting, “Google’s US$1bn Africa Milestone Puts Cloud and AI Infrastructure in Focus”, by examining what businesses must do after compute, cloud and connectivity become available.

Sources

Discover more from Towncrier Africa

Subscribe to get the latest posts sent to your email.

After AI Infrastructure, Who Owns the Workflow?

The digitalisation gap is really a usage gap

A model’s output is not a business outcome