Stop trying to boil the ocean - where to start with AI adoption

Your mileage will vary

As boardrooms scramble to define their AI strategies, technology leaders are finding themselves under immense pressure to adopt and integrate generative AI into their organisations. Much of this pressure is well founded. Today, AI is transforming how we live and work - a transformation that is only set to accelerate in the coming years. However, uncertainty remains, largely fed by extreme claims and counterclaims about its efficacy and ongoing potential.

Will future iterations of the technology live up to the marketing hype? Are we reaching the limits of the current wave of LLM technology? Have we reached the point of diminishing returns? How deeply can we integrate a non-deterministic technology into our DNA?

And therein lies the problem - how do leaders forge clear direction when there is so much uncertainty in the mix? Moreover, how do they shape a technology strategy around a technology that can deliver incredible results one moment but baffling inconsistencies the next. Organisations wishing to adopt AI must keep their eyes firmly open to the evident risks and limitations, not just the incredible potential.

When it comes to AI adoption, everyone’s mileage will vary. A company's ability to leverage generative AI will depend on context and how well existing workflows can absorb the risks (warts and all).

Fact from fiction

The challenges of AI adoption are numerous. From steering companies through seismic cultural change to managing stakeholder expectations, there are many questions to answer:

• Are we ready? Do we have the operational, technological and cultural maturity to adopt AI? Is our data in shape? Are our people prepared? Do we need to transform existing ways of working?

• Where’s the value? What are the key challenges that can only be solved by AI today? What will it cost to implement, and will we see a return on investment?

• Where do we start? Is adopting AI an all-in organisational transformation, or can we take a phased approach?

• What’s the endgame? Where is AI headed? Towards a fully autonomous agentic future, or will human-in-the-loop expertise remain critical? (Spoiler: it will, for now.)

Don’t try to boil the ocean

Too often, companies assume they can sprinkle AI magic over every aspect of their business and see immediate results. The reality is, the gap between where most organisations are today and where they need to be to take advantage of AI is so vast, that significant action is required to modernise before they can fully embrace AI.

Successful AI adoption isn’t about quick fixes. Rather, it’s about:

• Aligning AI initiatives with business goals.

• Upskilling teams to work alongside AI solutions.

• Getting your data in shape by making sure it is structured, accessible and AI-ready.

• Reconfiguring how you work to integrate AI meaningfully.

In other words, adopting AI is a significant investment, equivalent to any modernisation programme.

Start small - prove concepts

Of course, the only way to eat the elephant is one bite at a time. Instead of splattering AI over your entire real estate (e.g. by trying to automate complex, multi-stakeholder processes), start with low-risk, high-impact opportunities i.e. repetitive, labour-intensive tasks that are consuming valuable expert time.

This is where many businesses are today - validating the potential of AI, discovering its limitations and assessing the effort to turn basic proof-of-concept experiments into production-ready solutions.

Your goal with AI isn’t to replace human expertise but to augment and enhance it. We are some way off AI being capable of fully automating knowledge work, but it can significantly reduce the grunt work. Therefore, focus on workloads where AI can complement human expertise - repetitive tasks such as data extraction, document processing, boilerplate generation. This is where AI excels.

Humans, on the other hand, will be free to focus on higher-value activities such as problem-solving, strategy, and innovation. (That is, after double-checking any generated output.)

What’s your delta?

Not every proof of concept will evolve into a fully blown AI solution - that’s okay. The learnings will drive future direction and provide valuable insights into feasibility, risk and return on investment.

• Is the ROI compelling? If not, AI may not be the right solution.

• Is the use case worth the risk? AI isn’t suited for every challenge.

• Is the technology mature enough? Sometimes, the tools just aren’t there yet.

If an AI experiment proves promising, the next step is to assess the gap between concept and production. Do not be lulled into a false sense of the effort required to close this gap. Your team might have produced a compelling POC in a few weeks, but 80% of the effort will be spent completing the final 20%.

The challenges will likely start with your data, but success will depend on how autonomous and reliable your workflows need to be and how much human oversight can remain in the loop.

AI adoption is a journey

AI adoption must be viewed as a phased, strategic journey - one that requires careful planning, investment and iteration. Even a small proof of concept will have far-reaching consequences to how you operate and deliver value.

The organisations that succeed won’t be the ones trying to boil the ocean. They’ll be the ones that start small, experiment wisely and build AI capabilities step by step.

So where do you start? Simple: with one well-defined problem, a clear business case and a willingness to experiment, invest and learn.