Scaling AI Adoption in your org

One of the fun parts of my role is to tell stories. Stories of tech folklore but also often genesis of patterns. These patterns end up emerging as “best practices” in the long term. So here’s a short story of what happened with Eric Schmidt when he shared his office space with an engineer. Bear in mind this is 2001, Google was doing about $100M in revenue and still at-least 3yrs away from going public.

It was one of such co-sharing office spaces days. Eric (the then CEO) was speaking to VP of sales about revenue and asking why it looked low ~ 120M. Amit overheard the conversation and said its going to be 138M - much to Eric’s surprise. Eric later learnt, Amit had built an analytics engine to forecast revenue numbers. Eric watched revenue numbers inch closer to this number as time went on. As folks in sales, know predicting rev numbers have many far reaching benefits. The level of accuracy impressed Eric. He leveraged the platform and became its early adopters.

Now there are many lessons in this one story. The most relevant to this post is as follows. Great ideas are most effective when business and technology co-own initiatives. Also businesses start seeing value in tech when they see the impact first hand. This last bit - is the premise for the rest of the post.

Scaling AI Adoption in your org

Great ideas are most effective when business and technology co-own initiatives. Also businesses start seeing value in tech when they see the impact first hand. This last bit - is the premise for the rest of the post.

Often in my engagements with customers - one of the question that comes up most is as follows. How do we build the culture of innovation and tech adoption for ML and Analytics. The rest of the post focuses on the patterns that I have spotted where it has worked well.

These organisations have the following three very visible traits.

  • They understand that there is a strong element of cultural change involved
  • They move from a siloed to a cross functional based capability. This has elements of business, IT, analytics, training and operations all working together.
  • and they have a culture of failing fast. They are willing to try things even with only a minimal amount of confidence to succeed at the start

These companies look at adoption of AI as a people first problem. It starts with identifying the needs for up-skilling. Not only in the technology space but also in other functions like HR, Business, Traiing and Operations. Each function has its own set of learning and adoption to do.

  • HR would need to understand better which skills to look for - how to identify a good data scientist profile esp. since these roles haven’t been around for several years.
  • Business stakeholders need to fully appreciate the execution strategy of an AI implementation. This includes the principle of “human in the loop”. Increased awareness that business decisions won’t be auto pilot due to its implementation. Instead more informed decisions with the support of AI. Mature organisations start off with moving low/minimal risk, high toil workloads to auto pilot leveraging AI instead.
  • Operations also need to understand the impact - to re-consider workflow decisions.

These companies also build a portfolio of AI initiatives. Usually having each of the business units come up with their own. This ensures a widespread impact across the organisation, avoiding innovation in a silo. These initiatives have three distinct trends.

  • These initiatives are a good mix of long term and short term benefits. Those that need establishing foundations first and others that would reap benefits immediately. E.g. A retailer re-building their demand forecasting platform grounds up. They would need re-looking at the entire workflow. An immediate release of value could be something more short term. For e.g. building a strong AI driven recommendation engine. It would also data to feed into the forecasting systems as well in the future.
  • Staffing for these initiatives is such that it can be co-owned by both business and IT. This means success and failure are both owned by them. This ensures accurate mapping of current business process to corresponding technology process. Once the mapping completes, technologists can identify opportunities for automation and increasing accuracy. The teams also include training and User Experience. User experience teams help define how this change will affect its end users. Training teams build out a roll out plan. Having a plan ensures seamless adoption of every aspect of the change of workflow.
  • Clear measurement of value expected from the initiative. As real time measurement as possible. This allows for course corrections to pivot or speed up based on the outcomes. New adopters tend to keep this as low priority. As a result they miss out the early signals when things don’t go in the right direction.

The last aspect of this is organisational setup. Should this be a central team or a in-BU team. The practical answer is a hybrid model.

  • A central team can address cross cutting concerns. They can define standards, put in place data governance, recruitment strategies, training, principles.
  • The business unit specific teams can more specific to their area of capability. Areas like choosing priorities, operating model changes, implementation details and finally value realisation.

This model creates the required network effect in the organisation. This speeds up things next time a common problem shows up. Also provides the business units to make their own decision, improving agility.

How have you seen this play-out in your organisation? Have you seen great demos that never saw the day of light? What has been in your opinion the biggest challenge in the AI adoption in your organisation or team. I would love to hear more about your experiences.

Photo by Kevin Ku on Unsplash