By Hassan Qureshi and Matt Lemay
Which business activities can we accelerate or augment with machine learning that would impact outcomes of interest?
More business professionals are coming up with answers to the question above through visions of the possible prototypes, workflows and use cases. AI adoption is being increasingly mandated by leaders of corporations and departments.
As adoption rates are on the rise, the focus of local conferences and workshops has shifted from “What can AI do for us?” to “How do we deploy AI successfully?” This is due to a collective understanding that with this blue ocean of opportunities comes rough waves. We have seen many companies wash up on shore without successful or sustainable deliverables, while for others it has been smooth surfing to the success story.
Why is that?
Common issues in adoption
Across all the projects we’ve seen, failure (e.g., no long-term adoption) is largely attributable to these three main culprits:
- Poor data. Data may be overly sparse, unavailable, corrupted, largely irrelevant etc. Time should be spent up front to validate that data is complete, available, reliable and relevant for its intended use.
- Poor benefits realization. A strong benefits activity often involves benefits identification by the project sponsor; validation of drivers, assumptions and measures by the project manager; and periodic analysis of the progress towards achieving the benefits by an independent function such as quality assurance or risk management. Consideration of defined benefits and progress tracking should be front and centre during discussions on proposed scope changes.
- Poor ownership. A machine learning solution requires the same management and oversight as a new employee. Like an employee, it can be trained (i.e. supervised machine learning), monitored (i.e. audit logs), reviewed (i.e. validation of outputs) and promoted (e.g. enhancements to features and capabilities).
Getting ready to start
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.
Accordingly, when we have an exciting idea for an AI deployment, we do not need to race our competitors to the market. Rather, we suggest a risk-based-and-paced approach that focuses on addressing the common issues in adoption before development begins.
Specifically, we recommend a deep dive into the following questions:
- Data comprehension. How much data do you have and how usable is it? Is it limited by privacy or proprietary concerns?
- Process mapping. For each activity performed, is there a clear series of steps that leads to an outcome? Do we have an aligned understanding of how the AI solution will augment specific steps in the process?
- Benefits. After a process or activity is augmented or replaced, what are the expected benefits and how will we measure these (efficiency, risk mitigation, customer experience etc.)
- 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?
- 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?
Competitive advantage
We view AI projects as a 100-meter hurdle race, as opposed to a sprint. The hurdles are in plain sight and can be easily cleared with right technique, without adding too much time on our way to the finish line.
In our partnership, we have combined the risk management experience of MNP with the rapid deployment experience of Lemay.ai to carefully build AI solutions that are properly defined, supported and sustained through their lifecycle. In the past, we have developed several AI products that include capabilities such as trained AI classifiers that identify statements/ references of interest, and web scrapers that retrieve information of interest.
Ultimately, an AI project should create efficiencies in processes and/or augment employees’ knowledge and skill sets through new views and insights for decision-making. This is the competitive advantage that can be delivered through a sound AI strategy focused on establishing readiness prior to implementation.
Hassan (Hash) Qureshi, CPA, CMA, CRISC, CISSP, CRMA is a partner MNP’s Technology Solutions office located in Hintonburg. Hash and the 7 Hinton team provide advisory, digital and technology solutions and services to both public and private sector clients across North America.
Matt Lemay is an Ottawa-based serial entrepreneur. He is the co-founder of Lemay.ai (an AI consultancy), and the CEO and co-founder of Auditmap Technologies, an enterprise risk intelligence platform.
