Most artificial intelligence (AI) efforts fail. They don’t fail because of the tool, the core software, or bad data. They fail because they don’t integrate with business and wind up being more of a hindrance than a help.
This isn’t just an AI problem; it’s true of most forms of automation. Projects fail because the people building the solution have no clue about the actual goal, the nature and dependencies of their current operations, or even whether those operations are optimized. (In some ways that last one suggests an AI failure might be more beneficial – if you have a bad process in place, the last thing you want to do is speed it up!)
To achieve success, you first need to fix the process or operation, fully define it, set forth a set of achievable goals for the AI project and staff, and then execute on it. That’s why I’m fascinated by BCG, a consultancy that is increasingly focused on AI; its tactics evolved out of efforts to help companies improve operations with a strategic goal in mind.
Let’s talk about how to do AI projects right by using BCG’s focus on first fixing and optimizing whatever process you intend to improve with AI.
Speed vs. direction
When I was doing competitive analysis, I attended a lecture at the annual Society of Competitive Intelligence Professionals (SCIP) meeting that stuck with me. The speaker put up an X/Y chart showing speed vs. direction and argued that most companies focus on speed first — often resulting in a company going in the wrong direction faster. It seems obvious, but he argued that you need to be clear on the direction you want to take before you accelerate.
AI, and any form of automation, has a speed benefit. It can dramatically speed up whatever it is modifying. If it modifies a good practice, you get more good results. If it modifies a bad one, you get worse results faster, which can lead to a catastrophe.
The historic IT problem
The first time I experienced an example of this was when I specified one of the first CRM programs at IBM. When IT came back with the results, they not only failed to meet my specifications, they’d made virtually every problem I had more difficult. IT often fundamentally did not understand the process we were trying to automate and did not want input outside of the initial request. These problems were common. At times, things were so bad that there were running jokes about having to sacrifice chickens to get projects that did what they intended to do.
Since then, IT has become either better integrated with other areas or serves as more of a large-scale operations body — where software development or solutions can be created closer to, and often in, the operating units. (It’s also because operating units learned the technology.) With AI, because it’s still new, project teams tend to remain isolated and focus only on fast deployment. That brings us right back to that speed vs. direction issue: speed doesn’t assure direction or a quality outcome.
BCG demonstrates a best-practice approach. First, understand the nature of the process being automated with AI. Next, assure the optimization and effectiveness of that process. Then, after everyone understands the problem, the tools, the goals, and the best path to achieve them — that’s when you staff and specify a solution you can roll out. BCG also understands the skills you need for today’s blended human-AI partnerships; it can supply initial staffing while lining up qualified workers to assure success. (One recurring issue: if you don’t understand a technology, you don’t know what kinds of skills are best to support it.)
As a result, BCG’s AI projects rarely fail. That’s because BCG’s primary focus is not speed, it’s quality. BCG assures the result before it implements any AI solution, and clients get the results they want. (Note: BCG’s specialty areas are retail, transportation, healthcare, energy, industrial goods, and manufacturing; as with any consultancy, BCG will do best in the areas where its focused.)
AI is still emerging, and a lot of people who should know better are spending millions of dollars attempting — and failing — to successfully use it. It does not matter how cheap an effort like this is, if it fails, it was too expensive. And even if an AI implementation runs over-budget, if it succeeds, that overrun can be overlooked.
BCG’s focus on quality and understanding over speed — and the way it blends people and AI for higher productivity — could, if applied broadly, dramatically improve the success of these high-cost efforts. And that would deliver on the promise of AI rather than turning it into the train wreck it too often is.