Your AI Project Needs a Front Door.

Chances are your boss is or has been infatuated with what AI can do for her organisation. The infatuation might have evolved into a funded project to help solve a problem in your organisation.

The business problem to be solved could be one of many: predicting asset failure, classifying patient illnesses, forecasting demand, optimising routes, flagging unauthorised entry, detecting IP theft or any number of AI use cases.

But here’s the thing: 87% of AI projects never survive to deliver.

An 87% chance of failure.

Why so high? Many reasons are given. Perhaps the business problem wasn’t well defined, some say there’s too much of a technology focus and others blame organisational silos.

These are all plausible reasons, but these aren’t problems exclusive to AI projects. Why don’t design projects fail to the same extent? CRM or BI implementations? Or most other types of projects? Why AI?

Why 87%?

The invisibility curse

Having worked with a range of customers exploring AI solutions. There’s a common theme:

1. AI feels like a black box.
2. Organisational attention spans don’t cope with a three-month “drum roll” waiting for a model to predict data. Sponsors get bored and want to move onto something else that makes their success visible.
3. AI projects take a few months (or longer) to mature. After the honeymoon phase, people get antsy and ask, in frustration, “What is that team there really doing??”

These issues are unrelated to whether the organisation has clearly defined its business problem, or whether the organisation is operating in silos or not. Often the project sponsor is very clear on the outcome she expects. AI Engineers have enough self-awareness to call on business stakeholder input to help them model the problem.

More often than not, sponsors are under pressure to demonstrate tangible progress. something visible. Businesses move quickly, and by the time the AI project is ready, stakeholders have already moved on to the next shiny problem.

How do we keep stakeholders focused and feel like they’re working as part of the project, and not just waiting on it? How do we make them feel like they can contribute?

What if we made AI inclusive?

When we think of making AI visible, people jump to visualising the maths, variables and neural networks. This is sometimes useful, but it often adds to non-technical people’s sense of alienation, isolation and helplessness.

I speak with a lot of AI engineers, and there’s a common set of frustration:

• The business has no respect for the AI process
• The business thinks AI will solve the problem overnight
• Expectations are unrealistic
• They don’t appreciate the score our model generates
• They don’t see the elegance of our model

I was at a talk by Allie Miller (Amazon’s US Head of AI BD, Startups and Venture Capital) where she described user experience as one of the next frontiers for AI. And this struck a chord. It made me think: what if we helped the business experience AI in ways that helped people understand what we do.

Often the business isn’t funding the project to get a number, they’re looking to make the right recommendation. They’re not funding AI to get a model, it’s to simplify a complex process. They’re not funding AI for the sake of AI, they’re sponsoring it to improve their business.

In addition to asking “What is the number that sucks?” at the outset of a project, we also ask: “How do you want to experience the AI we create?”

Essentially, we help the business visualise how they experience our AI model, and get them to help us build that endpoint. This goes from being a black box project to one that requires multiple workstreams:

1. Creating the model
2. Process design in preparation for how we intercept the AI’s prediction and present it
3. Visualising the output of AI in some way

5 ways to give your AI Project a Front Door

1. Start with a picture.
2. Think about the action you want to drive off the back of your AI.
3. Express your AI as a visual experience
4. Make AI a feature of the business experience, not the business experience as a later phase.
5. Communicate how you’re moving towards Step 1 becoming reality.

The next few articles will expand on each of these phases with examples of how we do this for our customers at Sky Ledge. Subscribe to stay tuned!

About the Author

Ahmed is the CEO of Sky Ledge. He’s passionate about solving business problems using design thinking. Previously Ahmed held a range of roles across the University of a Director at the world’s largest Education Company, Pearson.

About Sky Ledge

Sky Ledge is ‘BX Platform’ that serves as a front-door for AI projects. We enable AI, ML and IoT teams to go beyond BI. Sky Ledge sends notifications about what’s important and enables everyday business users to consume the output of AI, ML or IoT services. Reach out on hi@skyledge.com