Be Bold With Your AI Projects – But Be Ready to Fail With Flair

It may have surprised an audience assembled to learn how to succeed with AI projects when ITV’s Lara Izlan told them to be prepared to “fail with flair.” But that observation received the biggest cheer of the day as it clearly resonated with an experienced crowd and was frequently repeated.

Everyone knows how difficult IT projects can be and the more complex the technology, the more challenging they become. It is why SoftServe brought together top industry decision-makers and IT talent at our first Executive Summit in London to hear experts discuss the best practises that can deliver successful outcomes when embracing artificial intelligence (AI).

During the keynote sessions, we heard about the importance of having the right data for AI to work, the need to align AI with business outcomes, the role of governance in making AI successful, and how technologists can win executive support and critical funding to proceed with AI implementation.

Izlan, who is the director of insights at the U.K.’s largest commercial television network, spoke of the ability of data and AI to deliver transformative business value when accompanied by the cultural and mindset change needed to be successful. Her early comments expressed a pragmatism born of experience that if you want to succeed, you must be prepared to fail. And, when you do fail, do it with a commitment to learning and trying again, or with “flair.”

Never give up

It became a common theme across the presentations. Embrace failure but try again. All agreed that AI projects, and those involving its more complex cousins Generative AI and agentic AI, are not easy. However, the audience was encouraged not to give up at the first, or even second or third, setback.

Disappointment is all part of a learning curve on the path to success. But we were told steps can be taken to minimise failures, learn from them, and then do things differently the next time.

The event was held at London’s Science Museum, which is not only a showcase of the Industrial Revolution but also the birth of technology that ensued. It houses early examples of computers and even AI developed by pioneers such as Charles Babbage and Alan Turing 80 years ago and more.

In an opening welcome, the museum’s Director Roger Highfield also challenged the audience to think about what they meant by artificial intelligence. He noted that science has still not yet even been able to agree on a precise definition just for “intelligence.” That concentrated minds from the start.

De-mystify

Several speakers had a go, particularly Andre du Preez, Head of Emerging Technology at specialist insurance broker McGill & Partners. He said the key to engaging business heads and decision-makers and getting their support was by de-mystifying AI and talking about it in a non-technical way.

Artificial intelligence, he said, was purely a mirror of human intelligence, but working at fantastically faster speeds. It can see, hear, learn, think, and act in ways that enable machines to mimic humans.

He used an analogy to explain the difference between a chatbot and an AI agent (agentic AI), with the first being akin to a hotel receptionist and the latter a concierge.

The receptionist/chatbot can answer simple questions it has been trained (pre-programmed) to respond to but is unable to deviate from the script. The concierge, or autonomous agent, however, learns from its interactions with guests and can begin to anticipate needs, develop its own script, and execute the action.

Stay in the game

Preez took this concept a stage further and said that before starting a prospective AI journey, participants need to decide which game they want to play. Was it finite or infinite? The first is like a sport with known players, known rules, fixed beginnings and endings, and referees, with a sole objective — to win.

An infinite game, however, will have known and unknown players, no fixed rules (though probably some regulation), no referees (but some regulators), and no defined start or finish, with the only objective being to continue playing.

Therefore, if we approach AI from a finite perspective, we do AI for the sake of AI. We deploy it in pockets across a business, aim for short-term wins and immediate ROI, and we might win. But, in an infinite game, we deploy AI to solve business problems and re-imagine how we work, aiming for firm-wide adoption that delivers value and long-term growth. More importantly, it keeps us in the game.

It is why, he said, AI projects only work if there is leadership from the top. If firms start with a business strategy, not a technology, and think big but start small. But most importantly — they start.

Data bedrock

Jon Hammant, the UKI Specialist Technology Lead for AWS dug a bit deeper, telling us that a firm’s data is the bedrock to create unique Gen AI experiences for their customers. He said they should start by using domain data to ground models, both structured and unstructured, and then decide how they want to use it with large language models (LLMs).

It is important to keep things simple to begin with, he continued, as it is possible to do many things that are quick and easy that will nevertheless still differentiate you from competitors. Firms can use proprietary data to extract value and embed choices that will deliver successful results. But only if they remember to first identify how both customers and the business benefit — not just one group.

Hammant also used an analogy to show how businesses should prepare for AI by comparing the process to the popular TV quiz show “Who Wants to be a Millionaire?” Use all your lifelines, he stressed. Start with 50:50, by narrowing down the scope of possible projects to keep it simple. Then, Ask the Audience — or talk to your peers in other industries to understand their experiences. Lastly, always Phone a Friend for help, which he likened to engaging specialist expert partners such as SoftServe.

He stated that in order to start your Gen AI journey, a firm should move fast to select the right use cases and appropriate datasets that can demonstrate value quickly. The next step should be to learn and organise by empowering teams through a variety of training opportunities. Finally, firms should aim for scale, but by building people, processes, and technology as that scale is achieved.

Business value

SoftServe’s VP for Digital Transformation, Gerard Welikala, told the room that while AI maturity is accelerating and adoption is increasing, business value often remains elusive. He supported that with evidence from our recent global survey by Wakefield Research, which showed only 22% of firms that have tackled AI could acknowledge enterprise-wide success with those projects.

He said the lag that is occurring in unlocking business value is partly because a lot of the investment in AI has been in technology and not in the business. He added that there is also often a failure to understand whether the desired ROI is qualitative or quantitative. He argued that to change this, AI needs to be used to tackle a business opportunity and to help people understand what’s in it for them.

Referencing Clive Humby’s famous quote that “Data is the new oil,” Welikala said that as AI — and agentic AI in particular — is fuelled by data, a lack of the right type of fuel will limit the value that agentic can unlock. It is equally important, therefore, that the data, because it is a raw resource, is appropriately refined and processed (cleansed) for it to deliver the desired results.

Once the foundations are established, he continued, to unlock the value from AI, firms must be bold with a focus on real business problems. They must plan to scale with a precise design, build a strategy, and establish governance from the outset. Finally, he warned firms not to forget the humans and ensure there is sufficient investment in education, training, and change management.

Deliver the promise

In her keynote presentation, Lara Izlan shared her key building blocks — the "three Ps" — for data and AI success: people, process, and platform.

Izlan acknowledged that most companies start by focusing on platform, putting in effort and investment to ensure that robust technology and capability foundations are in place to build on. She noted, however, that the technology must be adopted to deliver outcomes, hence the need to focus on people. This means ensuring you have the right operating model, diverse talent, and strong partners in place.

The right process then connects your technology capability to the business in partnership with end-users, so the technology can be actioned and deliver on the promise of the data. She stressed that inclusive ways of working and a willingness to experiment as important elements of a good process.

Calculated risks

During the wrap-up panel discussion, one of our experts highlighted that the AI world is moving quickly and getting faster.

Therefore, to remain competitive and successful in this environment, they said firms must take small, calculated risks now. They repeated the mantra to “think big, start small, but start somewhere.”

You do this by calculating small risks, then scaling and escalating. Try new things quickly, the audience was told, but if something fails, don’t be afraid to also kill it quickly and try something else. Keep working out where the value is, and where the ROI can be found. The answers will emerge.

Another said firms should also be wary of using “proxy metrics” when looking for KPIs and ROI, as there is a natural tendency for people to optimise the metrics they have been given, rather than tackling the underlying task. “You might start with the wrong KPIs. But just keep asking why you are doing this in the first place and the KPIs will emerge and evolve.”

Finally, one panellist reminded the audience not to forget the people. Although we are talking about AI and technology, in reality, we are still just trying to solve people’s problems. If we can use AI to make our people more productive, it will make them able to look after customers better, which in turn will deliver the results that will look after the shareholders. It’s then a win-win all round.