Human-machine symbiosis: a vision for AI adoption

For the past two years I have often spoken about creating a symbiosis between humans and artificial intelligence (AI). I see this as a state where AI does not replace humans. Instead, AI and humans focus on their respective strengths and collaborate on tasks to achieve a better overall outcome. Put too much emphasis on either side, AI or humans, and adoption will not deliver on its promise.

Last month that vision became a central theme in McKinsey's report, "The Symbiotic Enterprise: How Cognitive and Physical AI Are Reinventing Enterprise Execution" (June 2026). It is the first report by a renowned organisation to focus on the balance vision, and it gives a good insight into what they see as the levers to success.

The publication is a good moment to share my vision with a wider audience, using McKinsey's report as a validated proof point of how that vision becomes reality.

01 · Augmentation leads to mediocrity

Before the grander vision of symbiosis, let us look at how AI has been used to date, and why so many reports say organisations are either stuck at, or retiring, their pilots.

Augmentation and symbiosis are both forms of collaboration: neither replaces the human. The difference is what you optimise. Augmentation keeps your process and roles fixed and drops AI in to speed up the existing job. Symbiosis redesigns the process and re-allocates tasks to whoever (human or machine) is genuinely best at them, which usually moves people up into design and oversight while agents execute at scale. That re-partitioning is where the step-change comes from; assisting a fixed role can only take you so far.

AI has been around for longer, but GenAI really took off four years ago with ChatGPT. That makes the technology new, immature and unproven. When that is the case, smaller pilots make sense: they establish a view of how the technology works and where it fits before you move to bigger projects and changes. It also matches the position on Gartner's Hype Cycle, which says that at the peak you should be careful about making big investments.

This comfortably describes the current state, and it is backed up by the stats in the McKinsey report:

62 per cent of companies are experimenting with AI agents.
Less than 10 per cent are scaling those agents within any business area.
Augmentation of existing processes results in a 5 to 15 per cent productivity improvement.
Most deployments remain individual Copilot solutions, rather than larger multistep processes.

The reality is that we need to change tack. AI is developing far quicker than any technology before it, and as a result we can move past the pilot stage to production deployments. When we look at production deployments, we can learn from the past.

The software industry is no stranger to sub-optimisation. New software suites were often implemented without looking at the underlying process first, resulting in IT projects that did not deliver on their promise. The same applies to AI, but the negative impact of sub-optimisation is far larger. AI is faster, more intelligent and touches more processes, so any issue in its implementation is amplified too. The effects are larger. That is why we need to properly redesign the process to reap the full benefits.

Very few organizations report meaningful P&L impact because most still embed AI within existing workflows. (McKinsey)

This is exactly the point I have been making. Augmentation without reinvention is optimisation of a model that is itself becoming obsolete.

02 · Redesign works

The good news is that not everyone is stuck at pilot stage. There are real bright spots that show bigger redesigns producing exponentially better results.

Software development: shifting roles and tasks (Gysho)

Traditional augmentation: organisations report 5 to 15 per cent productivity gains when they support the existing team with code generation.

Redesigned model:

Humans focus on organisational analysis, process design and solution design, and supervise agentic coding flows where they approve and monitor progress.
AI takes functional design specs and turns them into tested, working applications, executing complex workflows and operating as teams.

Result: when Gysho deployed this method, we cut delivery cycles by 80 per cent while boosting the quality of the products we launch. A single human coder now scales to a team of more than 30 agentic coders.

Customer service: autonomous assistance delivered globally (Gysho)

Traditional augmentation: equip an existing customer service team with chat tools that generate responses and query data to answer customers directly.

Redesigned model:

AI runs all first-line support with specialised agents, dynamically switching between roles and escalating issues to humans when appropriate.
Humans ensure AI has the right knowledge and expertise to answer customer queries, monitor overall performance and respond to escalations.

Result: this model accelerated in-depth support for a smaller manufacturing company, delivering instant support across the globe regardless of time zone, and freeing up the core customer team to focus on service improvement.

Physical AI: warehouse and manufacturing transformation (McKinsey)

Amazon operates over one million robots globally. Its Sequoia system identifies and stores inbound inventory 75 per cent faster than previous robotic systems. Critically, Amazon reinvented the warehouse operating model around AI-driven orchestration, rather than simply replacing humans with robots.

Ocado runs thousands of robots simultaneously on a grid, coordinated by centralised AI. It reports near-perfect order accuracy and up to 99 per cent reliable, on-time dispatch.

Ocado shows that highly standardised physical throughput can trend towards very high autonomy. That is not a failure of symbiosis; it is a boundary condition. Symbiosis is strongest where judgement, proprietary expertise and relationships matter: exactly the terrain of IP-heavy firms. In commodity throughput, the human role shifts to system design, exception-handling and governance rather than disappearing.

Observations from the redesign cases

These early cases share the same themes. They produce exponentially better results than augmentation jobs, and they are not even pushing technological boundaries. They also show that neither AI nor humans achieve a good result without the other. That is the key: these cases succeed because they achieve a symbiosis where each role is assigned the tasks it excels in.

Humans:

Own the outcome and remain accountable for it, legally, ethically and commercially.
Govern performance, ethics and accountability.
Build trust and relationships.
Set strategic priorities.
Assess agent outputs and provide quality control.
Create, innovate and design (for now, though as we will see, the durable reason humans endure is not raw capability but accountability).

AI agents:

Execute complex workflows.
Read and apply enterprise knowledge.
Coordinate and iterate multistep tasks.
Orchestrate.

Each party brings unique skills to the table, and when organisations intentionally assign tasks to either, they see greater success. AI executes and scales; humans own, govern and oversee.

03 · Building symbiosis

That brings me to the vision I have held for the last two years: AI will transform the world as we know it, and the only way to adopt it successfully is to redesign work from the ground up, to achieve balanced collaboration between human and machine. That approach is valid now and will stay relevant for the foreseeable future.

We are well aware that some of the largest AI labs and Silicon Valley voices openly state their goal is to remove humans from the loop entirely. At Gysho, we reject that view. We believe the better path (commercially, ethically and in terms of sustained value) is to redesign work so that humans and machines each do what they do best. Displacement is not the goal; durable, governed collaboration is. We should not explicitly try to displace humans with AI; the two cannot succeed without each other.

A fair challenge: what happens when AI is good enough to do the "human" parts too, the strategy, the design, the judgement? The race to build ever more powerful systems is real, and it is the story that frightens people. My answer is that the human role endures for a different reason than raw capability.

Even where a machine could decide, we choose to keep a human accountable, because customers, regulators and society require someone to own the outcome, be trusted with it, and answer for it. That is a role no amount of model capability removes; it only makes it more valuable. Which is why symbiosis is the no-regrets strategy: if progress is gradual, you win on productivity today; if it is sudden, you are the organisation that already knows how to govern and orchestrate powerful AI. Either way, the discipline you build now is what lets you absorb whatever comes next, deliberately, rather than being swept along by it.

The other thing to appreciate is that the AI change will happen whether an organisation wants it or not. Most are seizing the opportunity, and any organisation that delays or waits will face a gap to its competitors that becomes an existential threat. So, just to stay relevant, AI adoption has to happen. And to make it a success, symbiosis is required.

McKinsey: dimension shifts

If we accept that the AI shift will happen regardless of our ambition to trigger it, how do we make sure our organisation can handle it? McKinsey's report includes a simple framework that structures the thinking and shows the axes along which leaders need to plan.

Dimension	Traditional enterprise	Symbiotic enterprise
Human role	Planning and executing tasks with AI assistance	Setting strategic priorities and supervising AI systems
Organisation	Fixed multilayer structure coordinating vertical functions	Dynamic flat network of outcome-orientated hybrid teams
Economics	Labour or physical capital as dominant cost drivers	Technology-intensive cost structure (variable, usage-driven)

That projection is a handy planning guide to ensure organisations do not skip any steps. Gysho's own internal processes went through the same transition, and show how those dimensions play out in real life.

How Gysho shifted dimensions

Our company thrives on innovation; it is central to our approach. We adopt early, disrupt internally, and automate to extremes. It keeps us lean and ahead of others, and it gives us the experience we need for client deployments. When we disrupted our development processes we moved large swathes of the coding process to agents, which is a good dimension-shift example.

Human role: our programmers and technical experts have moved up into supervising roles that manage up to 30 coding agents per person. Our management team still sets the strategy, but does not perform manual market studies or analyses; marketing agents do that continuously in the background. Sales is notified when our agents spot companies looking for our services, instead of performing outbound sales.
Organisation: our team has always been intentionally small, but it had hierarchy and distinct roles. Nowadays roles are more mixed, focusing on the task at hand instead of a role description.
Economics: we innovate, build and execute faster than ever before, and our team's capacity has tripled overnight. But the cost of technology has increased and becomes a fact of life. That cost is real; we cannot switch these systems off without affecting some of our core processes. We manage it actively, for example by using the right model for the job, so small tasks use a smaller, cheaper model. We also process some inference in-house, on our own devices. On top of that, we have alerts and hard stops in place, so if something uses far more tokens than planned, costs cannot escalate.

Organisational flattening

In its June 2026 report, McKinsey also mentions a flattening effect that takes place when agents are introduced. As we deploy more advanced, intelligent agents, they do both specialised, in-depth work and start managing that work. Our own agentic coding teams have management structures; they check each other. Layers of middle management become obsolete.

Beyond organising work, agents hand a wider group of people the capability to do more and understand it better. The same humans become more versatile and pick up a wider range of activities; AI does the mundane work while humans get more time for the activities they excel at.

Accelerated innovation

Finally, and McKinsey points this out too, innovation is going to accelerate. Agents give us capacity and scale that were simply out of reach in the past, at a fraction of the cost. With faster innovation, the organisation needs to adapt at an increasing rate. The rate of innovation and growth is less about human capital and more about technology.

Gysho experienced this first-hand. In our mission to deliver high-quality bespoke products at off-the-shelf pricing, we need to innovate across the delivery process. We innovated in a matter of months to create a fully AI-driven project process. Our portal runs and manages project administration; humans interact with the customer and supervise the AI's work.

04 · Hurdles and challenges to success

Of course, it is not all smooth sailing. AI brings as much risk as it carries opportunity, but I do not echo the doom scenarios circling around. I take a positive approach: we choose to be thoughtful about how we adopt, rather than letting it become destructive. Even so, it is worth being aware of the potential hurdles along the way.

Challenge	What it is	Impact & solution
Societal disruption & job loss	AI will disrupt jobs and how we work: it will displace jobs we have today and create new ones.	Needs direct management to transition existing teams into new roles and absorb the change.
Adoption failures	Doing AI as augmentation of current processes, or overreaching by deploying too many autonomous agents.	Benefits will not materialise, or damage is done by autonomous systems. Adopt thoughtfully and strive for symbiosis.
Vendor dependence	Committing to a single vendor ecosystem such as OpenAI or Anthropic, facing extremely deep lock-in.	AI integrates deeply into processes, and any lock-in can become a choke hold on the enterprise. Stay vendor-agnostic and use integration layers.
Cost runaway	AI deployed without regard for token cost or output value, resulting in runaway cost that exceeds human labour.	Rightsize models to the task, and assess cost per task for the use case you develop.
Leadership expertise gap	Enterprise leadership teams are not AI experts; they struggle to make the right call, or are distanced from enterprise-wide adoption.	CEOs need direct involvement in AI adoption, as it touches every aspect of the enterprise and changes it in a way never seen before.
Economic shifts	AI democratises expertise, accelerates innovation and kills economies of scale, changing how enterprises operate and compete.	Small organisations can compete, while established ones face more competition. Economic shifts must be deliberately adopted.

Societal concerns and disruption

The introduction of this technology will have a huge impact on society. During my time at Gartner, I attended a session led by analyst Mark McDonald in early 2012. He predicted the future path of technology and outlined how it would have an increasingly disruptive effect, eventually leading to a decade in which jobs are displaced and have to change.

Everything Mark McDonald said in that presentation has become reality. If that is a precursor for the future, the disruption is real and will last a decade, before landing in a new equilibrium where jobs have changed and the economy works differently, more effectively than before.

Getting ahead of that is key. Purposeful decisions on role splits, and finding the new balance, have the potential to shorten the disruptive period. If we let AI loose, or relinquish too much control to single parties, we set ourselves up for a risky outcome. But if we think about adoption, and purposefully design our transition and organisations around it, we can ensure it pays benefits.

05 · Moving to strategy and execution

A. Disrupt early and focused

One of our clients, the Total Negotiation Group (TNG), recognised that AI would cause an economic shift. As much as AI can be a threat, it can also be an opportunity. They saw that their unique IP was a perfect opportunity to create agentic systems for their customers, delivering their existing expertise in a new product and business model.

That early disruption is now paying dividends. As early disruptors, they now have an edge in the market, with a product that outperforms what their customers' internal teams are developing. Alongside their traditional business, their agentic Partner+ platform now places their expertise in customers' everyday processes, turning AI into a new revenue stream.

This is key: disrupt early and purposefully. It gives time to experiment, perfect and learn as the organisation works towards symbiosis.

B. Organisation before technology

Technology is merely supportive of the organisation and should not be a goal in itself; that is as true for AI as it was for any software before it. But organising for AI is even more important, because of the deep organisational integration.

Review the bottlenecks and find out what the real process looks like. Design a new setup that emphasises symbiosis, with a clear role for AI alongside humans. This means you will redesign processes and rewrite job descriptions, and some people may need training. The full design of a symbiotic process accounts for the new role split between technology and humans, which leads to a natural set of requirements for the wider project.

C. Measured and value-driven

Start with the end in mind. Clarify your corporate goals and how they are measured; any initiative you start must inevitably lead back to these priorities. The output of AI solutions must be measured in the same metrics, as a contribution to corporate goals in one of four categories:

Revenue growth: activities that lead to an increase in sales.
Market share: initiatives that extend competitive differentiation and grow market share.
Profitability: a focus on efficiency and cost to increase margin.
Compliance: activities that keep you compliant and within legal requirements.

In reality there will not be a single case; you will have many. Some suggested by your teams, others obvious bottlenecks. When you clearly map outcome metrics, you have a decision-making framework for what to do first. Balance decisions along two axes:

Feasibility: is it easy to achieve within the organisation and with current technological capabilities?
Value: what is the expected contribution to corporate goals, and how does the output measure against the required investment?

D. Balance operation with transformation

When you select initiatives, it is crucial to strike the right balance between operational use cases and those that will transform your organisation.

Operational: these initiatives change the way you work today, by redesigning processes and implementing the symbiosis between humans and AI.
Transformational: these initiatives change the type of organisation you are; they are large, multi-year projects that carry higher risk than operational changes.

As a general measure, organisations should spend about 80 per cent of their time and effort in the operational bracket, leaving 20 per cent for transformational projects. Both matter long term: operational changes help you achieve symbiosis early, while transformation gets you ready for economic shifts.

E. Choose future-proof technology

When it comes to choosing the right technology, staying flexible is key. AI is moving incredibly fast, and a technology or model you chose six months ago may be obsolete or retired today. There is also increasing concern about data residency, processing and AI legislation.

This matters in more ways than you might think. If you use models from Azure (or even Copilot as a tool), there is a good chance you are leaving the decision on model retirement and updates to Microsoft. If, as in Gysho's case, you run a platform with more than 68 prompt-engineered agents, any change has a huge impact.

On legislation, the technology you buy requires careful review. While Microsoft, Google and Amazon services are still widely accepted, the CLOUD Act conflicts with the EU's legislation, and the gap only widens as the EU AI Act comes into effect in August 2026. Depending on the type of data you handle, on whose behalf, and where you are headquartered, you may find technology choices are dictated by rule rather than preference.

Avoid lock-in: stay away from single ecosystems that lock you into one vendor. Price hikes will happen, and those who are locked in will foot the bill.
Achieve modularity: build modular stacks where you can exchange one solution for another without much work; technology develops fast, and newer, cheaper options will arrive.
Manage legislation: choose vendors that are compliant with local legislation and with the contracts you have with your customers. Do this early; migrating later is much more painful.
Consider bespoke: the cost of software development has come down rapidly, and bespoke builds are now viable for many more cases. Achieving symbiosis is much easier with bespoke solutions, and they are more likely to drive high returns.

This is not just theory. To put vendor-agnostic practice within reach of any team, Gysho has open-sourced LLMrPro, an intelligent LLM load balancer that routes requests across any provider, enables multi-provider failover and optimises cost in real time. Releasing the full codebase is our way of making sure developers and organisations keep the freedom to switch models and providers as the market evolves, rather than being trapped in the walled gardens forming around the major AI platforms. You can explore the repository on GitHub.

F. Build, fail and fix fast

Sometimes the best way forward is simply to start. That is very true for AI projects: as long as the organisational context is set correctly, the technology can adapt to facilitate it. On the projects Gysho has completed, these progressive insights materially change the final product for the better.

Start building a prototype to the spec, and experiment.
Invite teams to review and provide input.
Do not hold on to the original spec just because it is signed off.
Stay open to technology shifts and be ready to pivot.
Agree a hard deadline by which the product must be ready and live.

06 · How Gysho is positioned

For four years, Gysho has specialised in exactly what the McKinsey report describes: helping IP-heavy firms reinvent their operations through purposeful human-AI symbiosis. While others were selling AI augmentation tools, we were designing symbiotic operating models. While competitors optimised existing workflows, we were redesigning them from first principles.

The report validates what we have been practising:

Reinvention over augmentation: we do not layer AI onto broken processes; we redesign the processes themselves.
Balanced deployment: humans and AI each contribute according to their genuine strengths.
IP as competitive advantage: we help organisations codify their expertise into compounding agent skills.
Bespoke by necessity: off-the-shelf solutions cannot capture what makes your organisation unique.

For IP-heavy firms, this moment is both opportunity and urgency. Your proprietary knowledge, processes and relationships are the raw material for proprietary intelligence that compounds, the first source of durable differentiation identified in the report. But that advantage is only captured if you act now to codify it into intelligent systems before competitors do.

The symbiotic enterprise is not a future concept. It is emerging now. The report makes clear that this may become one of the largest sources of competitive advantage of the coming decade. Organisations that wait for clarity will find themselves competing against rivals who have already restructured their economics, redesigned their workflows, and deployed intelligent execution at scale.

The question is not whether this transformation will happen. The question is whether you will lead it or be disrupted by it.

At Gysho, we have been preparing for this moment for four years. We are ready to help IP-heavy firms navigate the transition deliberately: defining bold target states, building scalable foundations, and capturing the compounding value of proprietary intelligence. The symbiotic enterprise is here. The time to act is now.