More Human Not Less
How we’re using AI to do more of the work we love — not less.
AI is getting better fast, and in the last few months it’s started appearing inside almost everything we already use. We’ve seen Anthropic release its own relatively capable design tool, and even the tools we do our sensemaking work in (like Miro) now have AI layers built straight into them.
Increasingly we’ve been feeling the pressure to take a step back and form a critical point of view on how we should be using AI, before the tidal wave of agentic add-ons and automation layers washes away the core of our working culture.
It’s a sharper question for us than for most. Andthen is a small studio of experienced strategic practitioners, and between us we’ve run teams and directed plenty of other people’s work, but we’ve deliberately chosen not to build the studio that way. We structured Andthen so we’d keep our hands on the work itself — borrowing from Tom Critchlow’s thinking outlined in the strategic independent, where he discusses a sweet spot of being able to both work with executives on strategy, but also being unafraid to roll up your sleeves and dive into the on-the-ground work where needed.
AI threatens this, and we expect to see a gradual drift away from this way of working if we continue along the trajectory we see ahead.
In this post, we consolidate some of our current thinking about AI in our industry (high-touch strategic innovation consultancy) and describe our practical guidelines for intentional use of AI.
First, the bigger picture
Our main concerns about AI can be split into the system and the technology.
The System
We expect AI to reinforce the socio-economic system it’s being built on top of. That system is largely extractive — it runs on data, attention and labour scraped from people who rarely see the value flow back. It is instrumentalised — built to optimise, to capture, to convert. And by default it points AI towards becoming a tool for individual sovereignty rather than collective wellbeing.
The Technology
In the context of futures thinking, AI also narrows down our world of possibility. Futures thinking exists to escape from the tyranny of the probable. We’re often helping our partners think beyond the most likely future, to critically engage with what’s possible and within that what their specific, ideal future looks like. Generative models are probability engines, reaching for the consensus, average or mean. So regardless of whether they are used in futures thinking or not, they bias us towards a narrower option space.
The headline worry for us is inequality. The most likely trajectory is a quiet, familiar story in which AI accelerates every tendency the system already has — concentrating advantage, hollowing out the rungs people used to climb, and widening the gap between those who own the tools and those who are managed by them.
While these are our headline concerns, we’re going to park them here and explore them more fully another day. For now we want to stay close to home: what AI is doing, and could do, to our own work, and what we intend to do about it.
How AI could change our work
Here are the things we worry about in our own sphere of high-touch strategic innovation consulting. Plenty of these we’ve already to see good evidence for:
- The employability ladder gets pulled up
AI is very good at the lower-skill, repetitive work that juniors and graduates once learned the craft on. For a lean senior-led studio like ours that’s almost an advantage, but across the industry it’s a real problem: if the bottom rungs disappear, so do the pathways into the profession. - Insight gets flattened
Generic models all work across roughly the same context, and they all default towards the most common, obvious, average answer. Feed the same inputs to the same models and different practitioners will get similar outputs. The risk is a slow homogenising of insight — everyone arriving at the same unremarkable view. - We lose the writing-and-thinking muscle
Writing isn’t only how we communicate. It’s one of the main ways we think — the discipline of writing is one of the ways we work an idea out, find the holes, and make sense of something messy. Hand the writing away and you hand away one of your ways of thinking. - We lose our own voice
The more we lean on voice-to-text and on AI ‘tidying up’ our messy drafts, the harder it becomes to tell what’s authentically ours. Voice is a muscle too, and it atrophies. - We automate out the part of the job we love
For us, the joy lives in the middle of the work — wrestling with the complexity, getting unstuck in a sensemaking process, finally connecting two things we’ve been turning over for weeks. It doesn’t live in delivering seventeen outputs quickly. There’s a real risk of automating away the very thing that makes the work worth doing, and returning to a way of working where we don’t do any of the doing — we just manage AI to do it. - We reshape our work around AI
It’s easy to slip into a mindset where we’re forever bending tasks to fit the tool — “how do I do this with AI?” — instead of choosing the right approach first and finding the tool that fits. - We erode the relational core
So much of what we do is relational: understanding people, collaborating with each other and with clients, reading a room, building trust. It’s emotional intelligence and social observation. There’s a pull to let AI stand in for that — to make the judgement, give the opinion, offer the feedback, play the collaborator. - We transfer power to AI companies
We don’t believe our work is fully automatable (of course we don’t). But we do see a risk in becoming steadily more dependent on the handful of companies that provide these models — a slow transfer of power and self-reliance.
Our two north stars
In the backdrop of all of these risks, we want to build an intentional approach to using AI that focuses on the following two objectives:
- Clear out the busywork so we can focus on the core work
Use AI to take on the low-value, time-consuming, attention-sapping tasks — the ones with sharply diminishing returns on our energy — so more of our time and focus goes to the valuable, human, relational parts of the job. - Enhance our craft, but don’t let AI do it for us
AI should make the work we’re known for better — sharper thinking, stronger research, better relationships — but it should never stand in for it. The most important parts of the job stay ours and stay human-led. AI amplifies them; it doesn’t replace them.
The rules we’ve set ourselves
To hold those north stars in practice, a handful of plain rules.
- No first drafts
No AI first drafts of anything — especially of synthesis and sense-making. The first draft is where the thinking happens, so it has to be ours. It doesn’t need to be polished: a voice note, a sketch, a few bullets. It just needs to be ours. - AI doesn’t sense-make
AI can help with analysis and with organising what we’ve found — crunching quantitative data, finding correlations, summarising long documents. But deriving meaning, the actual sense-making, stays with us. - No synthetic inputs
We guard against generic insight at the input, not just the output, by feeding our work only high-quality material: primary research, ethnography, theory, lived experience — information that reflects the real world in all its complexity. AI doesn’t get to screen or select what feeds our synthesis until we have. - Nothing leaves unchecked
Nothing AI has touched leaves the studio without a human checking it. We — as individuals and as a team — are responsible for what we put our name to. - Two kinds of writing, protected differently
Writing to think (building a narrative or an argument) stays human, and often happens off-screen, on paper, post-its or a whiteboard. Writing to communicate, once the thinking’s done, is where we’re comfortable letting AI help. - Customise the tool to us, not us to the tool
Where it’s worth it, we shape AI around our context — training agents, building custom workflows, and, when it becomes feasible, training our own models for our specific needs. We bend the tool to the work, never the work to the tool.
A framework for deciding
Get Santini Basra’s stories in your inbox
Join Medium for free to get updates from this writer.
All of this comes down to a single correlation: the more core a task is to what we do, the less we let AI lead it.
In the bottom left sits low-value busywork. We’re happy for AI to lead this — near-autonomously, and in some cases fully.
In the middle is what we call the AI sandwich: a human starts the task and does the first draft or sets the parameters, AI delivers the middle, and a human checks the output — usually with a few iterative loops.
In the top right are the activities that are core to what we do. These stay human-led, with very little or no AI at all.
Above the corridor is the danger zone, where we’re letting AI replace core craft. Below is the inefficient zone, where we’re doing busywork by hand when AI could help. The aim is to stay on the diagonal: match the use of AI to the importance of the task.
What it looks like in practice
Here are some tangible examples of how this plays out in practice.
What success looks like
If we’re using AI the way we’ve described, we’d expect the following:
- We’re spending a greater proportion of our time on core work.
- We’re adapting AI tools to fit our contexts, rather than adapting ourselves to fit the tools.
- Our work still has a distinctive Andthen feel.
- We’re spending more time working collaboratively.
- We enjoy our work more.
- We’re limiting our dependence on AI companies.
Answering our own concerns
We started with a list of worries. Here’s how the approach answers each one.
None of this is finished, and some of these feel like half-measures that will need strengthening as we learn more. But, it’s how we’ve started, and this exercise has in particular been helpful to build a positive nuanced vision of Andthen’s future that we can use AI to reinforce.
