Nicholas Nesbitt | TeleSky Limited tenth anniversary | Nairobi | 9 October 2026
Celebrating a decade and asking a bigger question
Good morning to the TeleSky leadership, colleagues, clients, partners and friends. Congratulations on ten years of building a business, serving customers and investing in Kenyan talent.
An anniversary celebrates what you have achieved. It also gives you permission to ask a bigger question: what could we become?
Your theme brings together a decade of excellence, human talent and intelligent solutions. That combination matters. Intelligence without care can produce a very efficient bad experience. Talent without the right tools can struggle to deliver the service people deserve.
My own experience founding and leading KenCall gave me a deep passion for this industry. I believe Kenya can participate in the global economy through the ability of our people to solve other people’s problems. That belief remains strong. But the way we turn that ability into jobs and businesses is changing.
Here is the question I want us to carry through today: if intelligence becomes cheaper and more widely available, what becomes more valuable?
My answer is the ability to turn intelligence into something that works: a problem resolved, a business improved, a citizen helped, a life protected.
What AI changes and what remains uncertain
For much of the outsourcing industry’s history, our proposition was straightforward. We could provide capable people, reliable processes and good service at a competitive cost. Often, growth meant more contracts, more seats and more employees.
AI challenges that relationship. It can draft replies, summarise conversations, retrieve information and support decisions. Increasingly, AI agents can also take actions through connected software: update a record, arrange an appointment or initiate a transaction. A chatbot gives an answer. An agent may do something with that answer.
This changes teams. A smaller team may serve more customers. A supervisor may oversee people and automated workflows. An experienced agent may help define the tests that decide whether a digital agent is ready for customers.
We should be honest: some routine work will shrink. Some roles may disappear. Telling people that nothing will change would be disrespectful to them.
But a job contains many tasks, and automating one task does not automatically eliminate the whole job. In a published study of customer-support workers, access to an AI assistant increased issues resolved per hour by about fifteen per cent on average. Less experienced workers benefited especially. That is evidence from one setting, not a promised result for every centre.
The International Labour Organization’s 2025 assessment found that one in four workers globally was in an occupation with some exposure to generative AI. Its assessment emphasised transformation more than complete replacement. Exposure is not a forecast of job losses.
We cannot know the final balance. Technical capability, customer trust, regulation, cost and management choices will all influence it. We can prepare without pretending to predict everything.
Three perspectives that sharpen the debate
The conversations around AI, including those on The Diary of a CEO, are useful because they challenge comfortable assumptions. Three perspectives help me frame this moment.
Mo Gawdat, the former chief business officer at Google X, offers a serious warning. In Scary Smart, he argues that the way people develop and teach AI helps shape whether it benefits or harms us. In his August 2025 Diary of a CEO interview, he describes the possibility of severe disruption to work, power and society before a more abundant future.
Those are his forecasts, not an established timetable. We should not organise a business around a certainty that every job disappears on a particular date. But we should take his underlying challenge seriously: powerful technology does not automatically produce humane outcomes.
Daniel Priestley offers an entrepreneurial perspective. In his discussion of AI, he sees scope for small teams to build businesses that previously required much larger organisations. If the cost of creating and delivering a service falls, more people may be able to offer it and more customers may be able to afford it. That expansion is possible, not guaranteed.
Nick Hanauer adds an economic question: who receives the gains? On Pitchfork Economics, his discussions with economist David Autor explore whether AI can broaden expertise and strengthen opportunities for workers. Hanauer’s wider argument is that prosperity depends on a thriving middle class. More productivity alone does not ensure broadly shared prosperity.
Put those perspectives together. Gawdat asks us to take disruption and responsibility seriously. Priestley asks us to see entrepreneurial possibilities. Hanauer asks us to think about wages, demand and who benefits. My conclusion is that Kenya needs all three conversations.
TeleSky can help businesses make AI work
TeleSky’s own website already describes multichannel customer support, help-desk and IT support, alongside back-office and public-sector services. That is a useful foundation. The next opportunity is to help clients make AI useful inside the business, beyond an impressive demonstration.
Many businesses do not need another presentation explaining what AI is. They need someone to answer practical questions. Which process should we improve? Is our information accurate? How does the tool connect to our existing software? Who checks mistakes? What happens when it fails?
TeleSky could develop three connected services.
First, help clients adopt AI. Map a process, establish its current cost and performance, organise approved information, and select a manageable problem. Work with software and security partners to connect the solution to the client’s customer-management system, booking tool or other applications. Test it before expanding it.
Second, train the people who use it. A hotel receptionist, an insurance officer and a sales manager need different training. Teach them to describe a task clearly, provide relevant context, check the result and recognise when human judgement is required. Train managers to redesign work, rather than merely buy licences.
Third, run and improve the service. Maintain the knowledge base, monitor quality, handle exceptions, evaluate local languages and accents, and support staff when the system produces something doubtful. Businesses will need dependable operations after the launch celebration is over.
Imagine a distributor receiving customer orders across calls, email and messaging. AI could help turn an incoming request into a draft order. A trained person checks an unusual price, a stock conflict or a sensitive customer issue. The approved order enters the existing system, and the customer receives confirmation. The benefit is a reliable completed order, not simply a conversation with a bot.
The commercial model could combine a setup fee, a training fee and a monthly operating fee, with carefully defined performance incentives. For smaller firms, shared services could make this affordable. For larger firms, specialised teams could manage more complex workflows.
TeleSky does not have to invent a new foundational AI model. It needs the people, partnerships and discipline to make available technology deliver results.
The global opportunity and the changing shape of teams
This is also a global opportunity. A team in Nairobi could help a business elsewhere implement, test and operate AI-supported processes. But global reach is earned through competence and trust. Reliable connectivity, continuity plans, secure handling of data, relevant industry knowledge and the client’s legal requirements are part of the product.
The major BPO providers are already signalling this change. TP describes its direction in terms of combining human expertise with AI and automation. Concentrix has launched tools that assist human advisers and integrate with contact-centre platforms and business workflows. These are the companies’ strategies and product claims; they are not proof that every provider or worker will benefit.
The direction nevertheless matters. The contact centre can become a place that designs, operates and improves service, rather than simply supplying labour for individual calls.
Imagine a small delivery team: someone who understands the client’s business, someone who can configure the software, an experienced service professional, and someone responsible for quality and safety. AI helps them with analysis, drafting, documentation and routine processing. The team remains accountable for the result.
That creates possible paths for today’s staff into training, process design, quality evaluation, knowledge management and specialist customer support. Some paths will require technical qualifications. Others will grow from operational experience. None should be presented as a permanent shelter from further change.
If we train only for today’s script, we prepare people for work that may shrink. If we develop judgement, curiosity, industry knowledge and the ability to learn, we give them more options.
Kenya’s ambition should be to earn a larger share of the value we create, while building good jobs and credible career progression.
Government service should finish the journey
Now let us bring that ambition home. A citizen does not experience government as an organisational chart. A citizen experiences a missed appointment, an unanswered question, a permit delayed or an emergency that nobody seems to own.
A website can move a form online without solving the underlying problem. A contact centre can answer a call without delivering a service. The opportunity is to connect the request to an accountable response.
TeleSky could work with a county, an agency or a ministry on a defined citizen-service journey: explain the process, open a case, route it correctly, follow up, escalate delays and confirm whether the person actually received help. Repeated complaints can also reveal a process that needs fixing.
AI could support retrieval of approved information, translation, summarisation and classification. People would handle vulnerable citizens, disputed facts and exceptions. Access should include ordinary phones and relevant languages, not depend entirely on owning a smartphone.
How should government engage a centre like this? There are different models.
A competitively procured managed-service contract could pay TeleSky to run a defined helpline or case-management service against agreed standards. A shared service could support several participating agencies or counties, with clear ownership of each case. A funded pilot could establish whether the service works before any wider commitment.
A formal public-private partnership could be considered where a longer-term project requires private investment and meaningful allocation of delivery risks. Kenya’s PPP framework includes national and county authorities and recognises ICT and BPO among potential sectors. Suitability, affordability, feasibility, approvals and the lawful selection process still matter. A privately initiated proposal is a route to consideration, not an entitlement to a contract.
My recommendation is to begin with one problem, one accountable public authority and a measurable service. Choose the contracting model that fits. Success should mean cases resolved and citizens helped, alongside responsible use of public money.
An emergency centre can also create a local market
Consider a more ambitious example: an emergency coordination service reaching remote communities.
Kenya already has emergency numbers, including 999 and 112. The proposal would be to strengthen the response behind the numbers, in partnership with the responsible authorities and telecom operators. TeleSky could not simply take over a national emergency number.
Imagine a mother calling because her child needs urgent help. A trained dispatcher establishes the location and urgency under approved protocols, brings in clinical support when needed, identifies an appropriate available ambulance and confirms that the receiving facility can accept the patient. The case remains open until responsibility has been handed over.
That requires much more than a room full of phones. It requires trained responders, vehicles, communications, clinical governance, backup power, reliable funding and coordination with existing emergency services. AI could assist with records and logistics. It should not independently make a life-critical decision.
Now consider the entrepreneurial effect. A local operator may hesitate to invest in an ambulance because demand is unpredictable and collecting payment is difficult. A trusted coordination service could aggregate demand, dispatch qualified providers and support a predictable payment arrangement.
For remote areas, payment for readiness may matter as much as payment per trip. Otherwise, vehicles may concentrate where demand is densest. Government, insurers or another agreed purchaser would need to fund the model. An emergency patient’s ability to pay should not determine whether the call is answered or urgent help is dispatched.
Demand coordination alone would not make an ambulance business viable. But combined with standards, financing and dependable payment, it could help local entrepreneurs invest in vehicles, staff and maintenance.
Here is the bigger idea: a service centre can help organise a market. It can connect scattered needs to qualified local providers. The jobs created could include dispatchers in the centre, paramedics and drivers in the county, and maintenance businesses supporting them.
Agriculture disasters and public health
The same principle could support agriculture. A farmer calls about a crop problem, a buyer, transport or access to an extension officer. The centre records the need and connects it to a verified expert or service provider. Advice comes from approved agricultural sources, with escalation for uncertain or dangerous recommendations.
Aggregated requests might help an entrepreneur see enough demand to provide transport, irrigation repairs or other farm services. But farmers need transparency about prices and referrals. The centre should not exploit their trust to steer them towards whoever pays the largest commission.
For disasters, preparation is as important as answering calls after the damage. Before a flood season, a centre could help authorities update contacts, test escalation procedures, identify vulnerable communities and confirm available responders. It could disseminate official weather warnings and collect reports about blocked roads or rising water.
When discussing El Niño, we should distinguish preparedness for a possible climate-related hazard from a claim that an event is occurring today. Forecasts and public warnings must come from the relevant scientific and public authorities.
During an epidemic, a centre could explain official guidance, support appointment booking, route concerns to qualified health staff and follow up on authorised public-health actions. Aggregated reports could support surveillance, while health authorities validate signals and decide what they mean. A contact centre should not diagnose an outbreak from a rise in call volume.
Across these examples, the formula is similar: listen, verify, coordinate, escalate and close the loop. Technology helps. Institutions, resources and accountable people make the response real.
Trust and safety are part of the business opportunity
The more useful AI agents become, the more carefully we must manage what they can do.
An agent connected to email, customer records and payment systems may be able to expose private information or perform an unauthorised transaction. It does not need malicious intent to cause damage. A plausible wrong answer, an excessive permission or a malicious instruction hidden inside a document can be enough.
Security specialists call that last problem prompt injection: material the agent reads tries to redirect its behaviour. Imagine a message pretending to be an invoice that tells the agent to send customer records to a new address. The issue is not just whether the wording looks suspicious. The system must prevent unauthorised access and actions.
Individuals face related risks: persuasive scams, impersonation and exposure of personal information. A familiar-sounding voice is not sufficient proof of identity.
So give agents only the access they need. Keep clients’ data separate. Require independent verification and explicit approval for sensitive payments or disclosures. Test against manipulation. Maintain secure records of actions and a way to stop the service. Give customers a clear route to a person.
Data protection, including cross-border transfers, must be designed into the service. A Kenyan centre serving global clients must satisfy the relevant requirements for each engagement. That needs qualified legal and security support, not a generic promise that data is safe.
TeleSky could make this discipline part of its offer: helping clients deploy AI they can supervise, evaluate and trust. Human oversight must come with time, information and authority to intervene. A rushed person clicking approve is not an effective safeguard.
A practical next chapter
What could TeleSky do next?
Over the next ninety days, I would suggest three practical moves. Develop a small team drawn from operations, training, technology and quality. Give them protected learning time and a clear sponsor.
Choose one existing client process for an AI-assisted pilot. Record today’s performance, test against it, and measure quality, customer experience and total cost. Include software charges, review time and corrections. Scale only when the evidence supports it.
Alongside that, develop one public-service proposal with a willing authority: a defined problem, operating responsibilities, payment model and evaluation plan. An emergency dispatch project may require considerably more preparation. It should earn its way from feasibility to implementation.
Throughout this work, involve staff in redesigning the jobs they understand. Give people credible training opportunities. Measure career progression and job quality alongside commercial results.
We will discover opportunities that are difficult to name today. We will also try ideas that fail. The discipline is to learn early, contain the cost of failure and keep improving.
Closing and invitation to questions
Ten years ago, Telesky began building a business around human talent and service. Today, that foundation gives you something valuable to build upon.
The future may change how we define a team, a job and a business. Value may depend less on the number of people completing a task, and more on how reliably we solve a meaningful problem. But we still have choices about who participates and who benefits.
I want a Kenya that exports expertise, improves services at home and helps its people build better livelihoods. I want intelligent solutions that expand what human talent can achieve.
My challenge for Telesky’s next decade is this: identify a problem that matters, organise the people and technology to solve it, and make the result dependable.
Congratulations on your first ten years. Let us make the next ten years count.
Thank you. I would welcome your questions. And if we need a starting point: which problem should a centre like Telesky help Kenya solve first?
Questions
- If we stopped measuring success by seats and calls, what outcome would we measure instead?
- Which unresolved customer or citizen problem could we take responsibility for from request to completion?
- What should an AI agent never be allowed to do without independently verified authority?
- If our productivity rises, how should the gains be shared between customers, staff, owners and investment in new services?
- Which local industry could grow if a trusted centre aggregated demand and made payment dependable?
- What would it take for a caller in a remote county to receive the same accountable service as a caller in Nairobi?
- Which existing role could become a pathway into AI quality, training or process design, and what support would that person need?
- What new business could a team of four start now that it could not afford to start five years ago?
If AI halves the labour required for a contract should TeleSky try to preserve every seat?
Protect people through early training, transition support and new service development. Do not promise every task or seat remains. Explore whether lower costs unlock additional demand, and track whether actual new work is created. Workforce decisions should involve staff and consider livelihoods as well as margins.
Why would a client need TeleSky if it can buy AI directly?
Buying software is different from making a process work. The offer must cover integration, trusted information, staff adoption, quality monitoring and exception handling. If TeleSky adds no measurable value beyond the licence, the client has little reason to pay it.
Will small AI enabled teams create jobs or concentrate wealth?
Both outcomes are possible. Priestley highlights lower barriers to entrepreneurship; Hanauer focuses attention on the distribution of gains. Test whether new customers, businesses, wages and career paths grow. More output with fewer people is not, by itself, a job-creation strategy.
Is Mo Gawdat right that work as we know it will disappear?
He advances a severe-disruption scenario and questions whether new jobs will offset displacement. Treat that as a warning to prepare, not a settled forecast. His ethical argument in Scary Smart remains useful even if the timing or scale of his predictions proves wrong.
How would you get a government partnership started?
Find the authority responsible for a specific problem and agree the baseline, scope, funding and evaluation. Consider a lawful procured pilot or service contract first. Explore a formal PPP only when its investment, duration and risk allocation fit; seek current procurement and PPP advice before submitting.
Who pays for an emergency dispatch and ambulance network?
A defined purchaser such as a county, agency or insurer must fund it. A possible structure combines centre availability payments, ambulance readiness funding and verified trip payments. Establish affordability, payment reliability and rural coverage before vehicles are purchased. Do not make urgent dispatch dependent on a caller paying first.
Would a new 999 centre duplicate existing services ?
It could if poorly designed. Start by mapping existing command, dispatch and clinical systems. Propose integration or a specific coverage gap with the mandated authorities, not a competing emergency number or a parallel command chain.
Should an AI system decide which emergency caller is served first?
Any tool must operate within validated clinical protocols and authorised clinical governance. Life-critical triage needs qualified human responsibility, effective escalation and tested fallback arrangements. No pilot should remove an existing safe route to help.
Could AI agents steal money or leak information even when staff mean well?
Yes. Connected tools and excessive permissions can turn a wrong or manipulated instruction into an action. Restrict access, enforce permissions outside the model, separate client environments, independently verify sensitive requests, and log and monitor actions. Human approval is useful but not sufficient on its own.
How do we avoid selling another expensive pilot that never scales?
Choose a bounded process and pre-agree success and stop criteria. Compare against a real baseline; include review, correction, integration and software costs. Keep a workable fallback. Expand only when performance is sustained and a business owner accepts responsibility.
Can Nairobi genuinely serve international AI operations?
That is a credible opportunity, not an automatic entitlement. Win a specific sector or workflow through demonstrable results, reliable infrastructure, security, contracting discipline and applicable data-transfer safeguards. Start with a reference client and build from delivery.
What should a contact centre employee learn first ?
Start with safe use of an approved tool on a familiar work task: define the goal, provide suitable context, verify the result and escalate uncertainty. Add business-process and sector knowledge, then choose a pathway into quality, training, integration or specialist service. Access to learning should not depend only on personal spending.
