
19 minute read
SUCCESSFUL AI STRATEGIES FOR BUSINESSES IN TRADITIONAL SECTORS
A FRACTIONAL CDO’S ANTI-HYPE GUIDE TO DATA & AI FOR SMALL AND MEDIUM-SIZED ENTERPRISES
TONY SCOTT is a seasoned technology executive whose career ranges from software developer to Chief Information Officer. Beginning his journey as a consultant C/C++ developer at Logica, Tony built a deep technical foundation that led him into technology leadership roles at NatWest Bank, Conchango and EMC. Working as global enterprise architect, group digital transformation director and CIO at engineering companies Arup, Atkins and WSP, Tony championed data-driven decision-making. A consistent thread throughout Tony’s career has been his focus on leveraging data and emerging technologies, especially AI, to unlock business value, drive innovation and improve outcomes. With his focus on bridging technology and business strategy, Tony advocates for pragmatic digital transformation where technology serves clear business goals. Today, he continues to shape the future of the data-driven business as an advisor, fractional CxO, board member and speaker, helping organisations harness the power of data and AI for sustainable competitive advantage.
TONY SCOTT
Common Ai Pitfalls For Smes
What are the biggest mistakes traditional small and medium-sized companies make when implementing AI?
The biggest mistake is FOMO – fear of missing out. Companies rush into AI without a plan, pursuing ‘AI for AI’s sake’. They’re missing critical data foundations: proper infrastructure, clean data, and understanding of their existing information systems.
Many see AI as a magic wand, especially after ChatGPT’s launch. This hype-driven approach ignores essential elements like governance, security, privacy, and ethics – all crucial for successful AI implementation.
My mantra comes from lean startup methodology: think big, start small, scale quickly. You need the vision, but companies often jump straight to big-ticket items, which becomes costly.
The other major mistake is treating AI as a technology project instead of a business transformation. AI should deliver measurable business benefits and ROI. You need leadership buy-in, which is actually easier in SMEs than large organisations, but still essential.
Can you share examples of companies jumping into AI without a proper strategy?
the wrong problems – ones that don’t advance your business strategically.
The ‘garbage in, garbage out’ principle has never been more relevant. Companies need sufficient, clean, consistent data. They need to understand their system architecture – are data sources tightly or loosely coupled? Often, they’re tightly coupled, requiring data hubs, warehouses, or lakes to centralise and clean information before building AI on top.
Security and privacy are critical. Customer data must comply with GDPR, Cyber Essentials, and ISO 27001. Without addressing business drivers, KPIs, and outcomes, you’re just doing technology for technology’s sake.
The fundamental question: What’s your current data health? Do you have dashboards, alerts, and analytics? Are you using data effectively before considering AI? Your entire data estate – business systems, IT systems, operational systems – all contain data that needs proper storage, access, and querying capabilities.
Data duplication creates major issues. If customer address changes don’t propagate across all systems, you get inconsistent information.
I see ‘solutions looking for problems’ constantly since ChatGPT. A furniture manufacturer client kept insisting, ‘We want AI’ without defining their goals, KPIs, or business model. When I pressed deeper, they simply feared competitors had AI and felt they needed to catch up, but had no idea what to do with it.
A colleague worked with an e-commerce company desperate for an AI chatbot to solve customer problems. They spent heavily on implementation, but their real issue was order fulfilment. The chatbot actually made angry customers angrier. They eventually fixed the fulfilment problem – nothing to do with AI – and dropped the chatbot entirely.
Then there’s a flight training company with an impressive vision: using AI to connect student classroom behaviour with aircraft performance to improve training. Ambitious and transformative. But they had no data strategy, technology strategy, or IT foundation in place.
These failures result in wasted spending on tools, vendors, and staff for undefined problems. Projects get abandoned without delivering benefits. Since AI enables better human decision-making, building systems on faulty data produces wrong answers, eroding trust. The whole system can collapse like a house of cards.
The biggest missed opportunity? Using AI to solve
SMEs often use Power BI overlays or low-code/no-code solutions, but the core principle remains: clean, accessible data that drives business decisions and provides historical trends for AI prediction.
Ai Strategy Readiness Assessment
How do you assess an SME’s AI readiness in traditional sectors where digitisation might be limited?
I start with a leadership mindset. What’s their thinking around AI and data? Is it aligned to business strategy, or are they seeing it as a bandwagon? I always begin with business context first – what are their KPIs, objectives, and business outcomes?
A common thing I see is the desire to use AI for cost optimisation – very much bottom line focused. CFOs are looking down at the bottom line, wanting to automate and cut costs. But they should also look up at the top line for innovation, revenue growth, and newer business models. Really opening it to transformation.
I’m assessing their understanding and perhaps taking them on that journey to more ambitious but beneficial outcomes. If every company just looks down at the bottom line, cutting costs and automating, we’ve got a boring future ahead. Where humans can step in is around innovation.
The data estate and digital foundations all need evaluating upfront. There may be work needed before you can even consider AI. On the human side: leadership, culture, governance, ownership, compliance, depending on industry.
SMEs often think they’re too small for AI, but
TONY SCOTT
they’re sometimes surprised. There’ll be pockets where someone’s doing something quite advanced, perhaps not in their core role. You can gauge technical capability already.
I use a four-stage framework: digitise, organise, analyse, optimise – but in continuous loops. I’m currently working with a 30-person M&A company in London. From the outset, they understood they’re too small to automate people away. They want to make existing people more valuable, giving clients much better service. Very mature thinking.
What’s your framework for identifying high-impact AI use cases for SMEs?
I use a 90-day discovery phase about getting internal buy-in, momentum, and proving we can achieve value.
Month one is discovery and diagnosis. We sit down with key stakeholders, find their objectives, challenges, and decision-making processes. Are they data-driven or gut instinct? We do a data technology audit and map business objectives and goals.
Month two is prioritisation. From those business goals, we’ve probably got five to ten use cases –that’s the best number. We prioritise using a matrix: cost-benefit analysis, feasibility, and time to value. Hopefully, one or two key use cases come forward. We consider data privacy, regulatory issues, and risk, which might steer us away from particular use cases early. By month two’s end, you’ve picked a use case to pilot. Month three is designing that pilot. All metrics should be business metrics, success metrics, not about algorithms or model accuracy. It’s about reducing customer churn, machine downtime, whatever’s relevant to that industry. You get business ownership, buy-in, and authorisation to start work.
Everything must be defined in business language, especially budgeting. You want key stakeholders, including the CFO, involved. Look for relevant quick wins, measure ROI in business terms, but don’t have a big bang mentality – start small.
Define use cases in CFO-friendly language. You must score it and talk about measurable returns. Whether it’s customer churn, reducing delays, costs, or cycle times, it must generate genuine ROI where maintenance costs are less than benefits.
You also need to lay out a 24-months roadmap. Not fixed where you ‘arrive’ at AI, but ongoing because this world changes rapidly. You need quick wins upfront, proving investment value, but also take that longerterm view in the right direction.
Traditional leaders sometimes expect AI to be a magic wand. How do you help them understand that it requires investment like any business function? At the beginning, we talked about FOMO – companies rushing into AI with this sense of ‘we’re doing AI, we need to keep up’. But you should only do AI to solve real-world pain points and become a better organisation.
Some SMEs think they’re too small – only bigger players have the money. But I think it’s the opposite. SMEs can move faster than bigger competitors.
There’s a mindset about cost-cutting and automation. To bring your organisation along for the long-term journey, it needs to apply across the board –from middle management down to frontline staff. It’s about making existing staff more valuable. There are so many news stories about big organisations cutting thousands of jobs with AI as the culprit.
SMEs are leaner anyway and don’t have that ability. If they can make staff more productive, doing things where humans add value – automating tasks rather than roles – they can move much faster than bigger competitors and do more valuable things for clients.
Another mindset is ‘we need a data scientist’. There’s a rush to hire AI experts and ‘give us AI’. It doesn’t work that way. You need data foundations first, your current analytics approach, before you have the maturity to bring in data scientists. It’s a journey you work through.
Don’t see AI as just a technology implementation. Don’t tell your CTO, ‘Implement AI for us’. It needs true business leadership shaped in proper business terms.
AI FOUNDATIONS & PREREQUISITES
What data foundations and prerequisites must companies address before implementing AI? There are prerequisites – non-technical, cultural, and strategic. It’s what we were saying about not seeing AI purely for automating today’s ways of working, tomorrow. You can do that, but the much more exciting thing is augmenting current ways of working and freeing people up to give richer experiences internally or externally. Transforming your organisation into something new tomorrow.
If you’re solving problems with AI, you need to articulate those problems in business terms. You need ownership and accountability. Always having a business owner is really key.
The data piece – having clean, accessible, relevant data. The garbage in, garbage out principle. If AI is helping you make better, faster business decisions, you must trust the data. The data has to be clean first. A lot of organisations aren’t in that place already.
Just be sensible about team size or what you’ll have inside versus external partnering. Have expectations about setting up a function where you are today. You can be ambitious long-term, but know it’s a journey together.
What are the key differences in AI strategy between a 500-person company versus a 5,000-person company?
I’ve worked in 5,000-plus organisations and actually think it’s more challenging in larger organisations. They have multiple business units, potentially competing priorities, and different visions. You’ve got a C-suite sitting across all that, and it’s really key to get buy-in from every single C-suite member. It often becomes a business change. Sometimes the technology side is easy, but the human side and getting buy-in are much harder in bigger organisations.
They’ll have much more complex technology estates, possibly legacy systems going back years or decades. They’ve got more resources and larger technology teams. You can call upon internal teams, and they’ll be more specialised. Better ability to invest, fund experiments, and innovate. They’ll possibly have better access to large technology players – AWS, Microsoft – maybe account managers they can leverage.
Smaller companies are much more centralised and have more tactical use cases. Leadership may not be C-suite level, but it’s more informal and can sometimes move quicker. Technically, it’s a more cloud-first, plug-and-play approach. Internal teams are leaner but sometimes wear multiple hats – more generalists, actually more skilled in a sense. But they’re more cost- conscious, less open to experiments because they need quicker ROI.
In partnerships, they’re more open to bringing in partner companies to augment skills and provide knowledge transfer, hopefully forming strategic relationships.
To sum up, it can actually be easier. A smaller company will be more agile and, with the right approach, can perhaps move much faster.
REAL-WORLD EXAMPLE: WHEN STRATEGY ASSESSMENT CHANGES EVERYTHING
Can you share an example where your strategic assessment completely changed what a company thought they needed to do with AI?
Going back to that manufacturing company example, they got stuck right at the beginning, saying, ‘We want AI, we want AI’. They were very scared that competitors had AI and would outpace them without really understanding what that meant.
It was really about rolling all the way back, focusing on the KPIs, getting data foundations in place, and getting them in fit state before even beginning that AI journey. Then, really mapping out those AI use cases and ROI on each, linking them to actual true business objectives, which was a challenge. They had different areas with competing priorities, but it was really getting that leadership, that consistency of view and vision across the organisation. It was quite a journey.
We’ve been doing traditional data analytics for many years, and AI doesn’t invalidate those historical use cases – it just builds upon them. What’s the starting point? How does a company evaluate current performance? Things like dashboards and analytics –their relevance doesn’t go away because of AI. AI just takes you into more predictive worlds, doing more advanced stuff on top.
So in assessing maturity for AI, it’s always useful to see where the company is today in standard data analytics functions. If they don’t have that, it may make sense when sorting out the data estate to build those standard analytics functions before you even consider richer AI uses.
Can sophisticated, accurate reporting help organisations discover hidden business problems? Absolutely. In terms of reporting and analytics, there’s nothing new there, but some companies don’t have the infrastructure in place to do that well. You need the board asking questions about current business, current performance, and having that historical performance as well. I would always get foundations fixed first before rushing down the AI path.
Things like generative AI and large language models have given us new views on AI in recent years, but the same applies. If you’re training language models on internal data that’s not clean, outputs won’t be trustworthy.
I would fix today’s world – that view of the business today, standard analytics and reporting – then ultimately move toward a team that has all those elements. There are core foundational roles in data engineering – people doing actual plumbing of your data, making sure it’s clean. You need those in place first before you bring in analysts, then before you bring in data scientists on top.
LEADERSHIP CHALLENGES: BLIND SPOTS, CULTURAL RESISTANCE, AND STRATEGIC BALANCE
What blind spots do non-technical business leaders have when evaluating AI opportunities, and how do these lead to failures?
One would be what we’ve talked about – AI is all about automation and cost saving. It’s ignoring the benefits of augmentation and the human in the loop. My preference is to look for both. If you’re bringing your company and staff along with you, really look for opportunities to augment the mundane, rules-based, repetitive work they do to make them more valuable as people. There’s that human element – don’t just treat it as a technology implementation.
The other one is data complexity. Yes, how good is your data estate? But if you don’t have data engineers in place to manage that, it’s going to be very difficult moving forward. Don’t just say to your CTO, ‘AI is a technology project; implement AI’. It’s only about business outcomes. You’ve got to specify everything in terms of how it’s going to move the business forward. What’s the business return on investment? Have you prioritised things in a proper business way?
There’s a change management side to it. Don’t underestimate that – this applies to all digital transformation, not just AI. There’s a human element. You’re taking your organisation on a journey, so make sure you’re giving proper attention to change management.
How do you handle cultural resistance to AI in traditional sectors?
I mentioned earlier the news stories about threats to jobs. There’s genuine fear out there. In 2016, I was giving conference speeches in the engineering industry, making the case that automation is automating tasks, not roles. Yes, some people just want to cut costs and automate roles, but is that really the right decision? Look at every role and its tasks – you want to free people up to be more value-adding, things humans excel in: creativity, leadership, customer empathy, customer support. Free them from the burden of repetitive tasks rather than going straight down the path of cutting costs, cutting people.
To do that, you need to engage your staff, let them know your plans around AI, and assuming they’re not going to cut people, engage that front line early. Show them the quick wins you’re planning and how you’re going to improve their roles. Talk to them openly, honestly, and often. That phrase ‘human in the loop’ should really be the default.
Celebrate those wins and show people how AI is improving the internal environment. Try to remove that sense of it being a threat. A lot of organisations want to be AI first, but they should never be human last. That’s really a mantra and the cultural change management side of it.
Most organisations have company values reinforced in town halls, company meetings, and maybe posters around buildings. Just anchor everything you’re doing in those company values, so people see you’re not going against those.
AI companies talk about easy human replacement, but it should be more about enabling superior service delivery.
TONY SCOTT
How do you balance short-term wins with long-term AI strategy?
Both are important, and it is that balancing piece. You do need quick wins to set the scene and make the case. Without a long-term strategy, you’re not going in the right direction. If you just do quick wins, a lot of companies get stuck in endless loops of pilot after pilot after pilot. They build technical debt ultimately and fragmented solutions that don’t take them where they need to go.
On the other hand, if you just go for long-term transformation and forget the quick wins, people take their eye off the ball, you lose momentum, and potentially develop something that in 18 months is just not relevant. You did the wrong thing, or the market moved, or technology moved.
It is those quick wins that really give you proof of concept momentum. Perhaps you do experiments as you go to prove hypotheses, but you need that longterm anchor as well. That anchor is all about business value. Make sure everything you’re doing – be it a quick win or long-term strategy – is specified in terms of the business value you’re delivering.
Those quick wins are really the building blocks. We like to say the quick wins are the proof; the strategy is your plan for success.
BUILDING DATA & AI TEAMS: STRUCTURE, SKILLS, AND SECTOR CONSIDERATIONS
What’s the ideal structure for a data and AI team in a 100-500 person traditional company?
For smaller companies, I recommend a hub-andspoke model. They might not afford a full-time CDO, so a fractional CDO can bring experience from larger organisations and cross-industry insights while working part-time. Full-time is possible if the budget allows and you find the right person.
The hub needs one or two data engineers handling the ‘plumbing’ – data pipelines and infrastructure that stays in place long-term. That’s crucially important. Then I build data analysts or BI specialists as spokes, ideally embedded in business units. They may work part-time across departments, depending on how many you have. The key is keeping them close to the business so they understand daily dynamics, challenges, and problems. They answer business questions through data.
Only when you’re ready – when pipelines are solid, engineers are doing their jobs, and infrastructure is sorted out – do you bring in data scientists. They handle predictive models, simulations, preventive maintenance, customer insights, whatever’s relevant.
Sometimes they partner with technology companies for strategic skill augmentation on an ad-hoc basis.
How important is domain knowledge when hiring for data roles in SMEs?
There are bigger sectors where domain knowledge transfers well. Financial services have many elements – moving from insurance to banking is easier than jumping to completely different industries. Someone with an engineering career understands infrastructure, predictive maintenance, and can move between subsectors within that wider sector. But moving from engineering to financial services or health technology would be much more difficult. So yes, domain knowledge applies at bigger sector levels, but moving within them is probably easier.
What are the main differences between implementing AI in traditional sectors versus digital-first industries? Digital-first industries tend to be more data-driven. They’re newer industries built on data, so measuring performance and developing predictive solutions is inherently easier. They’re already thinking analytically before using AI and have a better analytics sense.
Traditional sectors can still make that journey –that’s where digital transformation comes in. But it’s a much harder ask, as much human as technology. You’ve got to convince them that historical working methods are less appropriate for the future. Senior management, middle management – it can be challenging bringing them on that journey. You’ve really got to start small, prove use cases, and do it non-threateningly.
There are real complexity differences, too. With digital solutions, you’ve got user data at hand and can easily measure performance and behaviour over time.
But I’ve worked on systems measuring traffic behaviour on motorways, tracking vehicles, predicting changes and traffic impacts. It’s much more challenging. You need many simulations, much more data, factoring in weather, climate, marrying together data sets that might never have been combined before. It’s more complex, but outcomes can be much more transformative.
I’ve done similar work with airports, mapping transport systems – road, rail – showing impact on arrival patterns and check-in queue lengths. Many systems that perhaps had never been considered for combined analysis.
The rewards can be quite transformative. It’s more creative, using design thinking and hackathons with business teams to answer questions we’ve never been able to answer before, or discover insights we didn’t even know existed in our data.

