Innovation with AI isn’t working

If data is regarded the new oil, AI is regarded the new electricity. Still, (or maybe because of that) according to research by Forbes, Gartner and DecsionsIQ, 85% of all AI projects fail. Why?

According to Zimmerman, among other things, expectations are just set too high. What AI projects seem to be doing is looking for really hard intellectual challenges and then trying to solve them with AI. At the same time, there are many more challenges that are less difficult to solve with better chance of good performance.

Another problem is that designers are often uninvited and unprepared.

What can we do to close the gap?

1. Dumb down the challenge: e.g., look for problems of scale rather than really difficult questions. Use the agile design process to look for an mvp with the first value, then build from there.
2. Include designers in the process; make sure there’s room for ideating different solutions, rather than jumping to implementation right away.
3. Work together fluently and interdisciplinary: designers, data scientists and domain experts.
4. Capture best practices: e.g. apply ux design in the double diamond process, but then also write AI design guidelines based on what you learned. Create new resources: e.g., create service blueprints with data layer lanes. Devise new processes: e.g., brainstorm on the basis of AI capabilities instead of what need to be solved.
5. Designers, prepare yourselves better: Get to know AI capabilities – e.g., spend some time with prefab AI’s

At Informaat we believe AI is the design material that designers must learn to utilize in order to create relevant products and services for tomorrow. Just like we learned what is possible with cardboard, wood, html, css or javascript. What do we have in the fridge? Well, if it hasn’t been sitting there too long, at least it is something edible 😉