Part 2: How to Design an Enterprise AI Workflow Capability Program

Most enterprise AI rollouts focus on training and prompts, leaving their people to figure out the rest. In the second part of this series, I outline a simple, scalable approach to building AI workflow capability across an entire organisation - including an experience to expose the skills gap and create the motivation to learn, tools and resources to close that gap, and a mechanism for capturing evidence to guarantee permanent transformation.

In the first part of this series Part 1: Why AI Will Fail Without Enterprise-wide Workflow Capability, we discussed the importance of building workflow capability before giving everyone AI i.e., helping people build the skill of defining how they do their work, which is a completely different skill from actually doing the work itself.

But how do we approach building capability across an entire organisation?

Well, this has been a perennial challenge for Learning & Development teams - the belief we must either:

  1. operate at an enterprise level e.g., “everyone needs training!” - which inevitably ends up with a generic knowledge dump that stays too theoretical to help anyone, or 

  2. focus on a specific use case and design a highly customised solution that only benefits one team e.g., “the sales team isn’t selling enough widgets, can you help?”.

Thankfully, with the approach outlined below, we can avoid both of these scenarios by giving people the skills and tools to improve their own work.

And this is the key here - we must create solutions that help people use AI within their own work, and not expect them to figure this out themselves.

Why AI training is doomed to fail

I’d like to start by crediting Nick Shackleton-Jones. In his book How People Learn, Nick talks about the insanity of building “courses”. His suggestion comes from the misguided belief that by cramming everything in a two-day event or three-hour online module, we’ll be equipped to perform better at work.

For those of us unfortunate enough to experience these courses, we know this to be a faulty assumption. We end up succeeding despite the course, rather than because of it.

Instead of courses, Nick suggests we focus on: 

  1. experiences to shift mindsets, and

  2. resources to help people perform.

He explains this in the context of ‘push’ versus ‘pull’. 

When we want to increase someone’s motivation, we should push an experience towards them (think about those pretend phishing emails sent out by enterprise cyber security teams that force people to think twice before clicking a potentially risky link because of the consequences of failing the test). The concept of this being ‘pushed’ is because the person may not be motivated to learn, so we need to push them into an experience that increases that motivation.

But when we want to improve someone’s performance, we should provide a resource they can ‘pull’ from at the moment of need (think about a handy tooltip in the software interface that shows people how the functionality works). The concept of this being that the employee is already motivated to act, and so they ‘pull’ from the resource at the moment it’s needed.

And so, building a best practice enterprise-wide workflow improvement program uses this exact approach:

1. Experiences: help people feel the problem

Our first job is to help people realise that they can’t leverage AI can’t without having first defined their work, which creates the motivation to master the skill. And then once they’ve acknowledged this, give them a chance to practice. 

And we can actually do this at the same time. Because what better way to demonstrate a skill gap than to experience that gap first hand? How often do we think “that looks easy” but when we try the thing, we fail abysmally?

So, by providing an activity in which people have an opportunity to define a workflow and a safe place to fail, we can help people see the gap in their own abilities.

And once they recognise this gap, their motivation to build that skill will increase.

I’ve found the most effective way to do this has been with an everyday workflow, something we’re all familiar with like making a cup of tea. And then asking people to deconstruct that workflow into a sequence of steps, providing enough detail that a novice could execute each step.

The mindset shift comes from the realisation that much of what we do comes from outside of conscious thought and is therefore missed when giving instruction to AI.

Once we’ve identified the skill deficit, we can give people tools to ensure that they don’t make the same mistake again. And then create space for them to use these new skills and tools on their own workflows.

The outcome being that once they have a well-defined workflow, the opportunity for how to leverage AI, automation or make other workflow improvements becomes immediately obvious.

And the beauty of this approach is that this experience doesn’t have to be delivered by a central learning function every time. It can be facilitated in a live session, run by teams themselves, allowing it to scale across an organisation.

2. Resources: help people improve their work

As part of the experience mentioned above, we can also give people resources to help them define their workflows: 

  1. to the correct level of granularity, and 

  2. with the correct level of detail.

A simple rubric can help with this (this is outlined in Part 5: The Easy Way to Define a Workflow to Leverage AI).

But once the workflow has been defined, we can also provide more advanced tools to guide people on how to improve their work (with or without AI). 

As we know, with enough context and instruction, AI can be extremely helpful. And so, with the defined workflow, AI can recommend how our work can be made easier, faster or better.

Because when it comes to using AI, the most common problem people have is not knowing how to use the technology in their specific use case. 

Why prompt libraries don’t work

As AI becomes the norm, most organisations are seeing a cottage industry of prompts:

  • “Try this prompt for writing a report”

  • “Here’s a prompt for summarising emails”

  • “Check out this prompt for creating spreadsheets”

And whilst these can provide inspiration, they also tends to create overwhelm and noise. 

A good comparison is the difference between the half-time commercials during a football match, versus tailored adverts in your social media feed. The former are generic whereas the latter are personalised and optimised for your specific tastes.

I imagine the conversion rates on these are wildly different.

The same dynamic exists within organisations. Help everyone and you actually end up helping no one (there will always be a subsection of early adopters who go the extra mile to figure out how to adapt AI into their own workflows. But the messy middle needs more help).

And so, we can provide tools to:

  1. help people define their workflows, and

  2. make recommendations on how they can use AI to improve their work. 

And by doing so, we’re not just giving people fish - we’re teaching them to catch fish for themselves.

Armed with a well-defined workflow and a tailored recommendation, many will go on to build their own AI solutions.

Evidence: How do we know it’s working?

Providing these tools also provides other significant benefits. Because once people are using them, we can capture data on:

  • the problems they’re looking to solve

  • which steps they find hardest, and 

  • where they want AI to help.

Aggregated across the organisation, this data becomes insight that leaders can act upon. Business units can deliver targeted solutions rather than continue the bombardment of generic AI advice, or top-down AI solutions that nobody uses.

Not to mention, teams building similar solutions can collaborate rather than working in silo, reducing duplicated effort. 

It also starts to answer the question every leader will eventually ask: is AI actually working?

Momentum: How do we make this a cultural shift?

Once people have built their solution, we can then encourage them to share their stories - the problem they faced, which solution they built and what changed as a result. Learning & Development can play a key role here, creating systems to surface the best stories and help them to be shared across the organisation.

Stories shift mindsets in the same way the experience does, but without needing a workshop for every person, and they funnel the next group of people in. 

Use creates stories, and stories drive use, creating a powerful flywheel that can drive a cultural change as workflow improvement becomes the default for how the organisation operates.

Thankfully, each of these pieces is simpler to put in place than it sounds. 

Over the next few posts, I’ll share what they look like in practice:

  • how to define a workflow in enough detail for a person or AI to execute it

  • how to design the workshop that gives people the chance to practise, and

  • how to build the self-service tool that guides them through it.

And if you’re an L&D professional who wants to master workflow improvement and become the person responsible for unlocking the value of AI in your organisation, join my email list today.

Next article in the series: Part 3: The Easy Way to Define a Workflow to Leverage AI.

Previous
Previous

Part 1: Why AI Will Fail Without Enterprise-wide Workflow Capability

Next
Next

Part 3: The Mindset Shift Needed To Build AI Workflow Capability