I was deep in a Miro board last week when I had one of those moments when something clicked (probably about two hours too late if I’m really honest with you).
I’d been sketching out high level org structure options for a client – thinking about the forms that their organisation could take and trying to get them beyond thinking about you the usual boxes and lines showing who reports to whom.
And I suddenly realised I had slipped back into designing for yesterday’s world. It’s like I was back on my first org design gig in 2010 again.
The options I’d drawn looked sensible enough on screen. But they completely missed the point of what actually creates value in my client’s business. So I had a black coffee, hit delete and started again.
Let me explain.
For the past eighteen months or so, I’ve worked on several organisation design projects with founders. AI has been a theme in all of them. Not the “let’s add a chatbot to customer service” kind of AI work.
The deeper question is really what does it actually look like to build an organisation that’s natively set up to exploit AI’s benefits?
And through that work I’ve come to a realisation that’s been nagging at me when designing organisations.
Workflow is at the heart of value creation.
I know, I know…that sounds obvious when you say it out loud. But bear with me, because the implications are more profound than they first appear.
I’ve been banging on about workflow for years – Team Topologies has been my go-to resource for a long time now. But what’s changed is this: when aspects of work in your workflow become substitutable with AI, workflow becomes a far more important design principle than it was even a few years ago.
In the old world, you designed your org chart around functions. Sales here, engineering there, operations over in that corner. That all dade sense because humans were the primary resource. And in most businesses, humans need to be organised into stable teams with clear reporting lines and career paths.
But what if the primary resource might not always be human?
Or more realistically, what if it’s a mix of humans, AI agents and automations that work in harmony depending on the task at hand?
I’m increasingly seeing scaling founders – or at least the switched-on ones – structure their businesses along the lines of their primary workflow. Not their functions. Their workflow.
This does something rather brilliant.
It reduces friction in the way work gets done and value gets created. But more importantly, it means you can substitute in and out AI-based agents or automations, or indeed add more humans into the mix where you really need them, all without disrupting the structure of the organisation.
Let me give you a real-world example.
One of my clients runs a B2B SaaS business. Their primary workflow is pretty standard:
- Find and connect with potential customer
- Qualify the lead
- Demo the product
- Negotiate and close the deal
- Onboard the customer
- Provide support and expand account through upselling
- Try not to lose the customer
In the old world, you’d have separate teams for each function. SDRs doing outreach. AEs running demos. Customer success handling onboarding. Support dealing with tickets. And so on.
But here’s what they’ve done instead.
They’ve organised as a matrix structure around the workflow itself. Each “stream” owns the entire customer journey for a particular vertical – from outreach through to account growth. Within that stream, they’ve got humans doing the bits that require judgment, empathy and creativity. And they’re continually looking at where AI can handle the more repetitive or mechanical bits – for example research, email sequences, data entry and basic support response drafting.
The key bit is this: when they discover that AI can now handle something that previously needed a human (or vice versa), they don’t need to reorganise the entire company. They just swap out one component of the workflow for another.
It’s a bit like building with Lego rather than carving from marble. Their org structure is modular, adaptable and designed for change rather than stability.
Let’s be honest, this matters because none of us really have any idea how quickly AI capabilities are going to evolve, despite what plenty of so-called “AI experts” tell us. None of us do. Anyone who tells you they know exactly what AI will be capable of in eighteen months is either lying or delusional.
As an aside, I’m reminded of the adage that we always over-estimate the impact of tech in the short term but under-estimate it in the longer term (it’s Amara’s Law if you want to quote it).
Anyway let’s get back to workflow-based organisation design.
So the question becomes: how do you design an organisation that can adapt as AI capabilities change, without constantly disrupting your entire structure and demoralising your team?
I think the answer is to design around workflow rather than function.
I’ll be the first to admit this isn’t a complete panacea. There are a few challenges:
Career paths become less clear. When you’re not organised by function, how do junior engineers progress to senior engineers? How does someone build deep expertise in a particular domain?
Hiring gets more complex. You’re not looking for “a salesperson” or “a customer success manager.” You’re looking for someone who can operate effectively within a particular workflow, which might require a broader or different skill set. None of us want to be trying to hire unicorns.
It requires more sophisticated coordination. When work flows horizontally through the organisation rather than vertically up and down hierarchies, you need different mechanisms for coordination and decision-making. These are the kind of practices that are vital for business operating systems to work properly.
But here’s the rub. These challenges are all second order things and solvable with the right focus. They’re design problems, not fundamental barriers.
For me the bigger risk is designing your organisation as if AI doesn’t exist. Or as if it’s just another tool that slots neatly into your existing structure.
Because it’s really not.
It’s a fundamentally different kind of resource. And it requires us to think differently about how we create value.
So if you’re thinking about your organisational design right now, here are a few questions worth sitting with:
- What is your primary workflow – the sequence of activities that creates value for your customers?
- Where in that workflow do you genuinely need human judgment, creativity or empathy?
- Where could AI (now or in the near future) substitute for human effort?
- Is your current structure optimised for workflow efficiency or is it optimised for functional control?
- If AI capabilities doubled in the next year, how much disruption would that cause to your current structure?
I don’t claim to have all the answers here. I’m working through this in real-time with my clients and we’re all learning as we go.
But I’m increasingly convinced that the companies that thrive in the next few years won’t be the ones with the best AI tools. They’ll be the ones whose organisational design is built to exploit those tools effectively.
And that starts with workflow.
best regards,
sw
From the Build archive
Build #64-
Organisational design that actually works
Focus on value, not fiefdoms when thinking about where to start with organisational design.
Build #42 –
The atomisation of work
A key trend affecting how we design organisations.
Build is Simon Wakeman’s newsletter for founders, ditching startup hype
to give you raw, practical wisdom about leading high growth businesses.
Simon helps founders navigate the transition from start-up to scale-up
through his work as a fractional COO, consultant COO, advisor and coach.
