VIEWS

Explore, experiment and execute: practical steps to implement AIExplore, experiment and execute: practical steps to implement AI

Now’s the time for AI plans to go from drawing board to development. Here’s a structured process to show you how to get there, and highlight the outcomes. Heads of digital are fielding versions of the same question from their senior leadership: what are we doing with AI? In some cases, it’s more like a mandate: start doing something with the technology. It’s no longer enough just to have a point of view; you also need a plan. Working with our AI partner, Phased AI, we take a user-focused view of the technology and we move towards implementation in a three-step process: from initial discovery of potential uses to a working proof of concept and then from idea to deployable software. We name these phases Explore, Experiment and Execute. Explore: a deep dive into data and decision making Explore is a research-driven engagement that usually lasts between four to six weeks. Together with Phased AI, the head of digital and their team, we learn more about the organisation’s customers, and about the structured and unstructured data it has, to see whether it’s a suitable building block for AI solutions. This guides the decisions to make about data they need, and whether it’s the right data from a privacy and ethical perspective. As part of this process, we spend time getting to know the would-be users and stakeholders who will be affected by this change. For AI to be effective and have value, it’s vital to understand this, to make sure the solution that gets designed or created actually solves problems for them. When this process is over, we create a map of a service, the whole business, or the customer experience. Then we draw where on the map we believe the AI opportunities are. We support that finding with customer research and technical research to say: If you solve this challenge using AI, it’s going to answer a real user problem and it’s high value. Targeting the right problems to solve When we say ‘high value’, we mean that AI isn’t just about solving a problem but about having a transformative impact on a business. Here’s one example: net promoter scores are well understood, but the customer service metrics they’re based on tend to rely on discrete data points. This leaves out a lot of unstructured data that hides customer sentiment. Here’s where generative AI is a game changer. It can delve into the subtleties of customer conversations, written reviews, and other types of feedback, and turn that into actionable intelligence. Not only does this augment the metrics you track today, it uncovers a more rounded and authentic picture of customer experiences and expectations. Put simply: the better a business can understand what its customers truly think, the stronger the connections it can forge with them. In turn, this drives strategic decisions with a depth and precision that were unattainable before. How AI enables new thinking AI, as a technology, enables a new way of thinking about things, so our design workshops are geared towards brainstorming to come up with new ideas or combining different ones. Here’s a real-world example: most people only read their insurance policy when they want to find out whether they’re covered for a specific scenario. That’s a perfect AI use case: what if you could save time instead of staying on hold to an agent? What if you can talk to your policy documents? Now imagine that as a mission statement for project owners or heads of digital to go to management or the board with the company’s AI roadmap for the next 24 or 48 months. All this work is predicated on collecting the right kinds of consent to unlock or to create these possibilities for your customers. This will often come down to a usability tradeoff: it’s why Google Maps has over a billion monthly users; most people don’t mind sharing their location, because they get a lot in return. Experiment: building on the opportunities Stage two, ‘Experiment’, takes the form of a three-to-five day session (which can happen in a single week or over several, based on your team’s availability). This sprint stage is where we design and visualise what that experience might look like and actually build a working proof of concept. This makes the technology very real very quickly. The third stage, ‘Execute’, involves taking the idea and turning it into a working, deployable piece of software. It’s worth emphasising, this is still not for external customers at this point, but a friendly group of stakeholders within the organisation. From execution to enthusiasm Execution is like a combination of the previous two stages; some organisations might go straight for this, depending on their AI maturity. This phase could involve some exploration, some research, an experiment, and several sprints. And then there’ll be a piece of technical work to realise the outputs of that experiment, and a sprint to the next level of detail. Everyone will be fired with enthusiasm at this point: there’s a live, working AI that people can see and interact with themselves and it looks amazing. So it’s important to be realistic here: turning AI concepts into software that’s potentially deployable to real humans involves huge amounts of technical work. This stage is an order of magnitude longer in duration than our initial project. In our experience, that’s true of anything to do with AI: it can feel very close when you start, but then once you actually get into the hard yards of delivering it, grappling with complex vector databases and retrieval algorithms, the endpoint can feel like it moves away from you. On top of the complexity, AI usually involves working with personal data and company data. Ensuring there are appropriate guardrails and protection around personal or sensitive information is critical – and, unfortunately, time-consuming. Setting the right expectations But now’s not the time to lose faith. Think back to how long it took your company to get digitally mature: it probably took years to assemble digital teams, build out your website, and put in place the infrastructure to support it. AI is similar: it’s a new phase in the digital revolution and it won’t happen overnight. While there’s still so much FOMO in the market around AI, it’s easy to make missteps. But with a guided process, you can emerge with a real plan for AI that has clear, identified benefits, a path to get there that avoids the hype, and with the advantage of technical expertise on tap. There’s still a marathon to run, and a carefully developed plan puts you firmly in the race. We’d love to work with you on moving your AI plans forward. See our AI service design and discovery page and get in touch.

by Chris Donnelly

Exploring UX Research in 2023

User Research is the foundation supporting UX as a discipline – if you're not doing research you're not doing UX.

But tightening budgets, murky ways to calculate ROI, and splintering of UX into several sub disciplines all point to a change in how businesses perceive the role and importance of UX Research.

Here's our round up of UX Research thinking for 2023.

Latest articles

Filter by:

Subscribe to our newsletter

Copyright © 2023 Each&Other Ltd.

Registered in Ireland. No. 545982

Privacy Policy