Implementing a Corporate AI Strategy

There is a cost to moving too slowly — almost as much as moving too fast

Published in

Towards Data Science

7 min readJan 19, 2019

In the wake of this generation’s digital transformation, machine learning and the greater promise of artificial intelligence creates wonder in people’s minds and effervescence within organizations. And the attraction to the field is justified: troves of process improvements are announced every day, every new tech startup has an AI play, and even governments are announcing their unique strategies to keep up.

The promise of fortunes to come is enough to give vertigo to the most grounded leaders. The break-neck speeds of innovation in AI means that new software ideas can be committed, checked out, trialed, inspected, verified and deployed within a few days. New features go step-by-step with continuous development cycles; new ideas with sudden business opportunities.

(For those that still doubt the speed at which this industry moves, here’s a story: we once designed, created, validated and deployed a proof-of-concept predictive model for maritime surveillance in an hour and a half — all we had to do is spend two weeks telling the client to go back and clean their data some more.)

Blockbuster in the Netflix era

Business activities and processes come and go; they’re intended to reflect the mood of the day, or capture an opportunity in time. Strategies, however, tend to move slowly and pervasively. They stick around after they’re no longer welcome, because “that’s how we’ve always done things.”

There’s a clear overlap between corporate strategy and corporate culture: innovation can be seen as shooting from the hip, just as deliberate planning can be described archaic and cumbersome. Having a deliberate approach to defining the problems your organization allows for people to make decisions independently, avoiding catastrophe.

What’s in a Corporate Strategy?

At the most basic level, a strategy is a “ high level plan to achieve one or more goals under conditions of uncertainty.” It is a series of heuristics and beliefs that is intended to provide an advantage (offensive or defensive) against an opponent (such as a competitor) or a situation (such as the market).

The point of having a strategy within your organization is similar to having a flashlight in a tunnel: you can’t make out the full picture, but you can be fairly confident in your next two steps, maybe three. And that’s all you need to keep moving forward.

Typically, corporate strategies can be summarized in a sentence or two and are somewhat obvious to the outside world — it’s the implementation experience of that strategy that’s proprietary and makes a firm successful. Deloitte is exploratory and flexible compared to KPMG’s buttoned-up procedural and reliable stance. IBM is losing its classic government and Fortune 500 advantage to new Cloud players like AWS and Azure because of their non-action on pricing plays and negotiation tactics.

Corporate Strategy Fundamentals

Back in 1996, Michael E. Porter wrote an extensive article to help define the guardrails around the “what” and the “why” of a corporate strategy. His takeaway points were the following:

Optimizing your processes alone will not drive more profits; your competitors are doing the same. Therefore, all companies in your industry will converge to the same optimal point.
To avoid this convergence and subsequent consequences, you can get an advantage by doing different things, or doing things differently.
However, the act of choosing what things to do next logically means that you are choosing to not do other things. Therefore, picking the right thing to do from a list of options suddenly becomes very important.

A good strategy will help support and justify your decisions when a situation is unclear, or a choice has to be made between two non-ideal outcomes. A bad strategy will breathe down your neck and coerce you into logically bad choices given the information presented to you.

AI Strategy in a Corporate Setting

For most businesses, the implementation of an AI play is internally-facing. This means that a good business strategy will be unique and differentiated from your competitors, but your AI strategy can be the same as everyone else’s. Your data is different from theirs, and therefore your results will be as well. A simple question to ask to launch this internally-driven strategy is the following:

Which business activities or processes could we or should we accelerate, augment, or replace with machine learning that would impact outcomes of interest?

To answer this question, there are a few inputs required:

Process mapping. For each activity performed, is there a clear series of steps that leads to an outcome? If not written down, do your team members have the ability to write down these processes?
Capability. Does your team have the full breadth of skills to implement a machine learning project? If not, what should be the focused role of an external team?
Measures of success. After a process or activity is augmented or replaced, what are the expected changes to happen?
Ongoing support. How will this new installation be supported over the next few years? Will there be an owner within the organization?
Scope. Is it just a product feature that needs to be implemented, or is there an entire department that requires an overhaul?

The execution and implementation of your AI strategy requires the same attention and focus as opening up a new department. Successful companies that have implemented and hired a data science team after our engagements gave it the same conditions as their IT, HR, or sales departments: with both a budget and a mandate.

AI Strategy Recommendations

There are additional considerations surrounding the adoption of AI within your organization.

First, a decree from the CEO is nice, but leadership buy-in at every level is even better. The typical resistance that we see is the fear of replacement or obsolescence from team members veiled as cynicism. Leaders across the organization have to be made aware of the focus on augmentation of staff, not replacement. There are even people that simply do not believe in machine learning because of the fanfare surrounding it — yet a sober process improvement that impacts the organization’s bottom line usually convinces them quickly.

Additionally, what’s more important than have having AI capabilities is having a data and analytics culture. If there’s no collection of results and outcomes, then no predictive or explanatory model can be derived. Machine learning needs to learn from data, and data comes from records, and records come from processes.

At the employee level, there are many considerations to support everyone involved. An AI project framework will help assist managers in providing a go/no-go decision. An in-house start-up culture will allow people to explore different approaches to a thorny problem. A continuous learning environment keeps everyone up to date. Having a single champion can be detrimental for morale if they leave, so ensure knowledge transfer within groups with show-and-tells.

Doing It Right — The First Time

In the larger organizations that we’ve consulted with, there was an explicit reason for them to bring us on board. They knew they had the resources to hire team of data scientists and let them run rampant, like hyenas sniffing out an injured gazelle; what they didn’t have was a second chance at building the team right as opposed to building the right team.

The recommendations we gave to them and the subsequent project implementations followed the following points:

AI should be boring. If it’s exotic, there’s a good to fair chance that it will be misunderstood, or even misused. You want a chef’s knife, not a blender. When there’s a clear problem, with a known solution out there, use it. Good is good enough when it comes to process optimization. Perfect never gets delivered, because there’s going to be a better research paper tomorrow anyways.
It directly impacts KPIs. Every endeavor should make your company more money, save your company money, or save your employees work and time. Clever plans fail. Go back to your organization’s fundamentals.
It will take time to implement. It needs to be implemented like a marathon, not a sprint. Remember that it took about 15 years for people to realize that an electric motor could be moved where it improved manufacturing instead of being slotted into its steam counterpart’s spot. Be patient, and focus on gaining momentum rather than overnight success.
It starts small. Most large endeavors fail. (In Canada, we’re still plagued by the echoes of the federal government’s payment system overhaul failure.) Dipping a toe in the water is safer than jumping off the deep end if you’re not sure if you know how to swim or not. Pick a project, break it down to its key activities, automate one of them, and then another, and so forth.

Final Notes

In his 2011 book Adapt: Why Success Always Starts with Failure, Tim Harford discusses the three errors that people tend to perform: slips, violations, and mistakes. A slip is pressing the wrong button, and a violation is when someone intentionally swindles you. As for the third category, the author explains it best. “Mistakes are things you do on purpose, but with unintended consequences, because your mental model of the world is wrong.”

In this new era of digital capability, the right mental model aligns machine learning techniques with business problems, and is ready to be reinvented tomorrow. Use this speed to your advantage rather than letting it be your downfall.

If you have additional questions about this article or our AI consulting framework, feel free to reach out by LinkedIn or by email.