Business Improvement with Data and AI Tools: A CEO’s perspective

AI tools are quickly evolving, did they impact your professional life as Managing Director, Chairman and Advisory Board Member?

Yes, absolutely. The most talked about use case for GenAI in particular is in coding, but I’m not a coder. Nonetheless, I’ve been using GenAI tools to help me in a number of areas and the amount I use them has increased steadily.

I’d say I’ve also refined how and where I use them during that period as I’ve learned where they genuinely help me. I do think there’s an element or trial and error required when using it as an individual, and you need to remain aware of their limitations.

Trial and error is more difficult when you’re thinking about using AI for core business processes and across large numbers of people, say for example in a law firm for contract drafting, and that’s where organisations need to be very careful.

Should business leaders consider AI tools relevant today? Can they bring sustainable value?

I do think it’s hugely relevant, but at the same time I’m conscious that a lot of companies have wasted time and money on things that haven’t been successful. It’s also not a one-way bet, you can’t just wave an AI magic wand and see benefits. Its use and deployment needs to be thoughtful and planned.

The basic use of ChatGPT or Copilot is now table stakes in a lot of businesses, so long as you put the right governance in place. It enables individual contributors in many roles to enhance their capability and productivity. This might be using it for background research, as a sounding board and thought partner around decision making or for creating basic creative content. As long as you’re not completely outsourcing your thinking and are checking the outputs, this is a fairly standard way to now operate.

At an organisational level, in terms of fundamentally helping measurably do things better, faster, cheaper that has been much more difficult for companies to deliver. Most of the best examples I have seen have been in one of three areas:

1. Process optimisation in operationally complex businesses (e.g. industrial manufacturing, logistics, warehousing)
2. Pricing and revenue management where these are complex and regularly changing (e.g. hotels, airlines, retail)
3. Information retrieval, synthesis and deployment in knowledge management (e.g. professional services, consulting, public sector)

Can you share simple, low-risk ways to implement AI processes or tools for companies?

For a lot of companies, I think there’s a fundamental question they need to first ask themselves, that will then define the nature and pace of their efforts with AI. Do you want to be a cutting edge innovator with AI in your space, with the risks associated with that, or do you want to be a fast follower, who takes the time to see which use cases really work, which software vendors really nail new products and then take and deploy the best of that?

The lowest risk ways are to empower people with enterprise versions of Copilot, Gemini, ChatGPT etc. and then, within a clear governance structure, get them using it. How widely you want to do this within an organisation depends on the sector, mix of roles and risk appetite.

The other relatively low risk way is to engage with some of the new AI-enabled SaaS tools that are targeted at your particular sector. This certainly isn’t risk free, as you still have all the challenges of successful SaaS deployment and the risk that new tools don’t succeed and aren’t around in 12 months time, but it’s worth looking at and potentially trialling something. Alternatively, many of the SaaS tools you are already paying for, such as your CRM, Accounting or HR system may well have AI enabled features that you may find helpful.

What are the most common mistakes companies make when trying to implement AI projects?

I see a few consistent mistakes.

Firstly, delegating AI to the IT or CTO function as if it’s primarily a technology issue, or forming an internal AI committee of people interested in AI and letting them drive this. Those groups will be important in the deployment of what you do, but ultimately the thinking around where and how to use AI should, I believe, come from the CEO, with input from commercial and operational functional leaders. CEOs need to be involved in defining how ambitious they want to be with AI, and most importantly where in the business it can have the most impact, and what the ROI on that looks like. This can be difficult, as most CEOs don’t have a background in this space, but they need to bring the right internal and external people around them and set the direction, as they should be doing with other core aspects of strategy.

Data is crucial to most AI projects, and I’ve seen two issues here. The first one is not taking the time to check that the required data is actually available to enable the AI use case – I’ve seen otherwise well scoped projects fall over immediately when it turned out that crucial data simply wasn’t available.

The second one, ironically, is the reverse issue where companies think that all of their data has to be perfect before they do anything with AI and spend months working to perfect a data lake across all data sources. If you have the right data engineers working with you, that degree of perfection is rarely needed for individual use cases and waiting for it can slow projects unnecessarily. So data is vital, but not every piece of data in the business needs to be perfect before you get started.

The final issue is launching into building custom solutions without thinking about how something can be productionised to work reliably in the real world. This includes how you’re going to manage a custom solution when aspects of your business change. If you build 100 custom GPTs (Generative Pre-trained Transformer) for people, what then happens when underlying models change and the tools start to underperform? Or how do you update your AI pricing tool when your inventory changes or customer behaviour changes?

Can project management best practices be applied to projects that include AI processes and/or tools? Should teams adapt the methodology to improve results?

I think a lot of the project management techniques used in the pre GenAI era are still entirely relevant. There are some new courses that have been developed specifically for AI project management like AIPM or CPMAI, which I think are also very good and a great place to go for those looking to heavily engage with AI projects. It’s easy to forget that the best AI project in the world can still fail if those people who are deploying the tool, and there are always still people involved, aren’t bought into the project and don’t change their behaviour. So I think engaging with prospective users before, during and after the project is more important than ever.

An additional challenge of projects involving AI is that the underlying technologies and their capabilities are changing so rapidly, that by the end of a long project the way you’ve done something may already be suboptimal. So, I think project management practices that allow a degree of flexibility and agility are required, and build your solutions in ways that individual components can be changed relatively easily in the future as technology moves on.

Is there any practical advice you would like to give to companies so they can prioritise without falling behind?

I think the starting point has to be to ensure that, however simple it is, you have an AI strategy in place and one that you continue to review and refine over time. Then remain curious, interested and up to date on what is happening with AI. Ensure that it remains a live topic in leadership meetings and under regular review. Even if you’re holding off on jumping into this space for now, there will come a time when you’re ready to do so and the cost-benefits of doing so make sense. You’ll have done your homework, seen what good looks like in your sector, and be ready hopefully to get it right the first time!