People
AI is no longer new news. It is disrupting industries and opening up opportunities for new levels of efficiency and effectiveness. Executive respondents to the recent AlixPartners Digital Disruption Survey validate this, with 75% asserting that AI will be extremely important to their industry.
However, many organisations still grapple with how to approach AI – just 21% of respondents gave their companies an ‘A grade’ in how well they are leveraging it.
The typical questions and challenges that we hear include: “Where should we start?”; “We’ve spun up a number of MVPs, but they are not having an impact operationally”; “Is the right strategy for us to move first, or be a fast follower?”; “How do we manage the risks?”; and “How much will this cost?”
GenAI has captured the imagination more recently, but many of the success factors are similar for more classical AI applications such as machine learning. We see five hurdles that repeatedly get in the way, and recommend how successful organisations can overcome these to gain serious traction and business value with AI, whether generative or otherwise.
1. Use case selection
The recent hype around GenAI may be subsiding, but businesses are under pressure not to fall behind. Fear of missing out can inspire interesting use cases that don’t deliver bottom line benefits, or lead to a proliferation of use cases that dilute the focus needed to succeed.
We believe the starting point must be the business goals and strategy. Focus on a manageable number of key opportunities and challenges where AI can make a big difference. Start small, learn quickly and be ready to pivot – once a use case is succeeding, scale fast and focus on making it fully operational.
2. People and change management
AI will only lead to business benefit if people are willing to adopt a new technology and change how they work. The “human” element of AI application and adoption should not be underestimated; change management at every level of the organisation is critical.
Successful teams will have strong sponsorship from the CEO, clarity on the business outcome, a product owner who involves the business users in the design, testing, and iteration of products – all working alongside the technical data and engineering experts.
Many organisations are successfully allowing colleagues to experiment, find AI productivity hacks, and scale them. This requires clear guardrails and guidelines around what is and is not acceptable but, done well, this fosters familiarity with AI and encourages adoption.
3. The gap to operationalise
Many organisations get to proof of concept (POC), but then fail to deliver on the promise by fully operationalising. POCs are often very successful initially, but require greater effort to be scalable and sustainable – highlighting the importance of focusing on only a few use cases.
To operationalise, consider systematic solutions to address data quality and pipelines, ensure the technology is secure, and develop new ways of working so that solutions will continue to be optimised. Regulatory considerations must be covered, and edge cases will need to be ironed out. This might initially take longer than you’d like, but it will accelerate – and pairing external expertise with internal teams can help.
4. Building on sand
One of the critical mantras in the world of AI is that it will only be as good as the data it can access and the quality of those assets. Many organisations realise that their data strategy is unfit for purpose, and that data capture and management needs an overhaul. This needn’t slow down progress – successful organisations are resolving and building their data capabilities while delivering powerful use cases.
Beyond the data itself, strong foundations must extend to the technology architecture, governance, and programme management. Skills and capabilities are also key – building the capabilities and teams needed to develop and adopt AI as it comes online.
5. Not being smart about risks
The risks associated with AI are very real and need to be taken seriously. Some “traditional” risks are now amplified (e.g., data privacy, cyber risks, and bias), and some new risks have emerged, such as hallucination and unintended consequences. Set against this, of course, is the risk of standing still and failing to deliver the business potential.
We recommend a pragmatic approach. Develop basic policies and ensure that everyone understands them; keep a register of data and AI assets, so you know what you have, how it works and what it is being used for; and ensure that a named human will always be accountable for every outcome. Then, embed risk considerations throughout the design and build process, take conscious choices around which risks are and aren’t acceptable, and build in mitigation where needed. The level of sophistication can develop over time, but a thoughtful approach from the start will reap dividends.
Conclusion
Amid the continued excitement surrounding AI, maintaining a practical mindset is critical. Firmly focus on the specific business objectives that you want to achieve, and see through the key initiatives.
Don’t overlook the data, technology and organisational foundations, and work hard on the change management – realising the benefits of AI might require a fundamental recalibration of core business processes and structures.
Understanding how AI can help your business is only the beginning. With the right tech-minded leadership approach – including advisors and partners to complement the knowledge base already on board – the subsequent journey can be designed and delivered with the strongest chance of delivering that business value.
Look out for our Practical AI for CEOs playbook, launching later this month – an essential guide for business leaders on how to approach, implement, and strategically leverage AI to maximise business value.
