Doug Greenhough is Head of Learning Tech at Impact
How to AI
We’ve all heard the hype. We’re variously excited, nervous, cautious, dismissive, optimistic. We’ve played around with ChatGPT and MidJourney, we use AI tools… when we remember. Maybe taken it a few steps further with properly integrated AI tools as part of our productivity suites, CRMs and other internal platforms? Enthusiastic early adopters have probably copy/pasted this blog into an AI tool and asked for a summary already, and will thus miss out on all the terrible jokes hidden within. But for most leaders, the challenges and possibilities around AI are still far more of an unknown than a known problem.
This blog won’t tell you what tools to use or where to use them, for a couple of reasons: Firstly, the landscape is changing so fast that any recommendations I make here will be out of date by the time you can implement them. New models, new tools are arriving at a hectic pace, and the performance and capabilities of tools on the market is constantly evolving. Secondly, despite having 20 years of background in tech, a solid grasp of the principles behind the current wave of innovation, and spending a lot of time working with different tools over the past two years, I have no idea what will work for you in your organisation. What I can offer you over the next 1667 words is a roadmap to a destination that you have to choose, and some signposts to important stops to make along the way.
Constructive constraints
The first pit stop we need to make when presented with any new tool is choosing where and how to use it. This is such an easy step with most software that we don’t even consciously make it, because most tools have a pre-built use case – a CRM, accounts suite or video production tool – and we simply decide if we need to do that thing, and if that tool is the best suited one for our needs. They’re screwdrivers; we know we need them to turn screws, and we can work out what type of screw needs what driver.
"Think of AI as an enthusiastic intern"
The challenge that we face with AI is that it fundamentally isn’t like that. It’s a paint brush, not a screwdriver, and what we use it for is far more open ended. Don’t get me wrong: there’s some very useful hyper-specific implementations out there, there are a lot of incredibly general-purpose non-AI tools too, but the current wave of AI is around more general purpose, language-based generative tools. A slightly pessimistic take is that they are calculators, but for words, and having a calculator doesn’t make you a mathematician any more than having a paint brush makes you a great artist. A more optimistic (and more human) view is that they are a very promising and enthusiastic intern; you can give them almost any task and they will give it a good try, but they lack some of the experience and critical thinking that they need to be truly great, and require good leadership and supervision to thrive.
How to overcome AI implementation challenges
In the face of this wide-open problem space, a common approach is to just… start. Do something – anything – learn and iterate. I usually really recommend that as a great approach for small scale experimentation, but where it commonly falls down is in extrapolating those early successes into a wider business context. You do the experiment amongst a small team (who are likely picked because they’re proficient and enthusiastic), everything goes well, you roll it out to a wider team… and suddenly there’s loads of push back, low adoption and poor performance. That’s unfortunately pretty common with all change projects, especially technological change, but for some reason implementing AI stirs up much stronger feelings than something like changing email platforms. When we reflect on what went wrong we’re likely to identify a lot of symptoms: poor business readiness, lack of consultation, an unexpected shortfall in tech literacy. These are all genuine concerns, but they aren’t the core of the issue: the real problem is that we didn’t stop to consider the existing culture into which we were launching this innovation.
So we come to the title of this section: constructive constraints. The first step when you have that blank canvas of infinite possibilities is to define the kinds of problems you’re looking to solve, and I’d argue that the best way to do that is to consider it from a cultural perspective.
Align your use of AI to your cultural perspective
What does your organisation really care about? If it’s efficiency, then using AI to automate repetitive tasks is likely to be well received. If it’s spending time building relationships with clients, then trying to save time by automating communications is never going to fly, regardless of how effective it could be. Target the tasks that the business doesn’t value, and the people doing them don’t enjoy.
Constructive constraints let you reign in that infinite problem space by confining them to the area between things you will always do, and things you will never do. Where those lines are will be different for every business, and should be a discussion at the most senior levels of business. As a life-long technologist I can say with absolutely certainty that it isn’t the tech department’s job to set those kinds of business-defining cultural cornerstones in a vacuum, and any attempt to do so will end in disaster.
Once you have those constraints, what kinds of task are left for you? Scheduling, taking meeting notes, interpreting reports, filling out RFPs? A good place to start might be to think about the counter-cultural things that happen in your business; the things that are necessary but nauseating.
"What necessary but nauseating tasks can you use AI for?"
Armed with your list of unpleasant tasks, consider how you could automate them, and to what extent. Obviously you can’t blithely send out AI generated contracts without opening a huge can of potential liability worms, but maybe a GPT trained on your existing contracts could do a decent first draft that’d save time better spent negotiating the details. Maybe you decide that you want all your marketing to have an authentically human voice, but AI can be useful in ABX testing your website. Perhaps you decide that an imperfect AI translation into hundreds of languages is better than no translation at all.
Practical AI applications in the workplace
All these considerations should go in to identifying the biggest thing you can fix, that people want fixed the most. That’s where you should start if you want to be successful, but it’s only the first step. Even culturally-aligned change fails if the people who have to change aren’t prepared for it, or feel like it is being imposed. For your implementation of AI to be successful it has to come from a culture of innovation.
Innovation culture
More ink has been spilled on this topic than HP sells in a year, so I’m going to keep this brief: innovation is not about you. I have led business innovations that have been incredibly successful and ones that have fallen completely flat. The difference isn’t the quality of the idea, it’s the quality of the people who make it happen. You need people who are capable of leading and being lead, of trusting others and being trusted, of empathy and openness, inventiveness and rigour. That is a lot to ask of people, and frankly not something you can recruit; even someone with all of those qualities will wither in a culture that doesn’t reinforce this mindset.
"There's no formula to create an innovation culture"
There’s no hard and fast formula to follow to create an innovation culture – we’re all starting from different places and with different ingredients – but some of the most critical factors common to all really innovative teams appear in Impact’s leadership action model.
Every member of a team must be aware of internal factors (personal and team dynamics) and alert to external forces (market and wider organisation needs). They must have the appropriate context and be trusted to make meaningful decisions. They need to be supported when they choose to act on those decisions, and need the space to reflect on that action and how it might inform future choices. Coincidentally, those are all points lifted from Impact’s model of leadership action; a force of change generated from everywhere in a team or organisation.
Adopting this model of distributed leadership can be a challenge for team members used to more hierarchical control (and to leaders), but it’s essential to fostering a genuine culture of innovation.
Leading change with AI
So you have defined your constraints and identified ways to implement AI, you’ve put the effort in to building a culture of innovation by empowering independent leadership action… what’s left for those of us in a traditional, capital-L leadership role to do? Are we redundant in this new model? Maybe we should investigate how to replace ourselves with AI? Unfortunately, no. We’ve made our job harder.
Integrating AI and leadership
Leading teams that are empowered to think for themselves, innovate and act is very different than motivating people to deliver a single vision. Everyone pulling in their own direction would be chaos, but our job is not to squash that chaotic energy with externally imposed order; it’s to guide it in a useful direction. Think about those constraints you defined earlier: the things your organisation always does, and never does. Imagine them like the outside walls of a fibre optic cable, and you as the wave guide taking the random bursts of energy and gently nudging them into alignment. Not only do you need the same small-L leadership skills we’ve already considered as key to innovation, but you need to be an exemplar of them to the people around you. An example of how you want people to behave: considerate, aware, strategic, confident, reflective. Your positional authority is so much less important than how well you reflect the values that you value, and how you nurture them in others.
"AI will rock some people's worlds in ways that they don't enjoy"
There’s a huge amount of fear and uncertainty around AI. Some people who really should know better are predicting that it’ll be the end of the world (as we know it), but the truth is that it will rock some people’s worlds in ways that they don’t enjoy. Creating a culture that takes change in its stride, and providing empathetic, human leadership isn’t just necessary for implementing AI successfully, but a moral requirement of making the attempt.
Harnessing AI for business success
If you’ve read this whole article then hah – I got you. You were searching for “How to AI” and I’ve dumped a whole bunch of philosophy of leadership (and even a pretty niche physics metaphor) into your brain. But that’s exactly my point: you can’t implement AI properly if you can’t land change, and you can’t land change without empowering the people who have to “do work differently, or do different work”. These are the same hard to learn, hard to teach skills that Impact have been helping people develop for 40+ years, but they are exactly what’s needed right now to effectively leverage this most modern disruptor into a new strength. Doing all that requires both a holistic view of the organisation in its environment, and that unmistakable, irreductible human quality of connecting with individuals. If you can do that, then implementing any tool, including AI, becomes a simple extension of what you already are, not an impassible edifice or a mire of contradictions that it can at first appear.