The AI Advantage - 2026

It is my belief that 2026 will be a groundbreaking year for AI in the enterprise. I believe this to be the case for two reasons:

• Enterprises are coming out of their discovery and testing the water phases that have been bubbling along with more or less fervour for the last two years. Paced by the capability of the models and product offerings from the new AI companies.

• The aforementioned “frontier” AI companies, primarily OpenAI, Anthropic and Google, are all seriously targeting the enterprise. Looking to push more enterprise ready products onto the market. This will create more competition and more tools and services for the enterprise.

Towards the end of 2025 OpenAI’s CEO Sam Altman stated that selling to the Enterprise would be a priority for the company going into 2026. Anthropic have already positioned their products squarely in the Enterprise space. Google is positioning itself as a safe, secure and open partner for enterprises.

Google continues to ramp up flexing its established enterprise credentials. Pivoting the entire organisation to an ‘AI first’ footing with a large portfolio of new AI products and AI functionality being injected into existing products.

All of this makes for an some exciting and maybe some scary moments to come in 2026.

But how will this manifest inside of organisations in 2026? What opportunities will these capabilities unlock for established organisations and new organisations? I think the impact will be broad.

The AI Opportunity

Throughout my career I have worked for small, medium and large organisations all at different stages in the organisational lifecycle. From what I would call an ’established start-up’ to massive multi-nationals.

I think it is a truth that generally the smaller, closer to the start-up end of the spectrum a company is the more agile and current or modern the company can be or appears to be. However in the start-up phase, while companies don’t carry as much technical debt which enables the agility, they of course lack the valuable body of knowledge that more established organisations have built. They also lack the enterprise guardrails. If you have been in a company growing just beyond the ’established startup’ phase you might have seen the growing pains that come along with having to try to avoid trading agility for operational rigour.

On the flip side, within a large established organisation (which I want to focus on in this article), legacy and a burden of technical debt from many years of running a successful enterprise creates a drag on agility. This results in a sense of lagging behind the innovators in start-up land. Many times I have seen the response to this to be for the larger organisation to try to replicate what the smaller up-start is doing. However, it is often the case that the engine of an established business may not react well to change or modernisation attempts that are modelled on start-up behaviours. Established systems running an established business can’t move quickly and break (to much) stuff.

Where am I going with this? Well for the start-up, enterprise AI offerings accelerate a go to market proposition. I think the use case here is clear. “Take and idea from notebook to shippable product quicker”. In 2026 ideas will indeed go from page to product much quicker. Ideas will perhaps be more valuable and initial execution commoditised. But more than just this, AI tooling and capability used smartly within a well run start-up can be employed to smooth out those growing pains. There is an opportunity to avoid completely the need for the traditional bureaucracy required to run a ‘grown-up’ enterprise. Start-ups that deploy AI capability smartly, beyond just vibe-coding a prototype will be able to go beyond an MVP much quicker. In turn, narrowing the capability gap between challenger and incumbent much quicker.

For established enterprises the calculus is different. Established organisations have a few things that start-ups lack which can be super powered with the use of AI: A body of knowledge and (shock!) a legacy. There are blunt ways of using AI tools to attempt to derive immediate value at a most basic level. But smart organisations will look beyond the basic.

Information Gold Panning

I am sure its a familiar story to many who work in or have worked in large organisations, there always exists large bodies of disjoint and disconnected, unstructured data stored in multiple different repositories built up over time. For even the most naive deployment of existing enterprise AI tools this is a treasure trove and grist to the mill for a machine tuned to ingest and process data.

The ability to quickly wire a natural language query interface on top of organisation data repositories has been transformative. The ability of the tooling to draw out institutional knowledge and capture fresh insight and value is transformative in and of itself. Yes garbage in gets you garbage out, but the tooling can act as a marvelous garbage filter. Or perhaps a super efficient Gold Pan that can find the nuggets of value in the slurry of accumulated data. The amount of time I saved myself in 2025 simply collating data across multiple data sources has been a huge win. The fact that it is technically quite simple to do feels like magic.

This is just a basic enabler. It gains you greater access to your existing knowledge. But it doesn’t intrinsically help you use it any more than before.

Super Power

Technical Debt. There I said it. Again. The drag on a large established enterprise that can make it feel like a slowly moving behemoth compared to the nimble challengers. But what if you could turn it into an asset?

In my experience tech debt gets out of control for two reasons:

• “If it ain’t broke don’t fix it.” If your thing is working well and making money for you then why rock the boat? However if renovation is left too to long, you get stuck on older tech, miss a paradigm shift and have to build new things in compromised ways. Before you know it you are stuck in the trap.

• “We don’t have the people or the time” Usually after getting trapped by “If it ain’t broke…”. There is another trap here though. Not spending the time addressing it makes it worse and costs you more time in the long run. It’s the terrible cycle.

We have strategies for dealing with this. Such as measuring value not just in terms of features that drive new revenue, but also in terms of protection of existing revenue such that the value of addressing tech debt is not lost. Or Using error budgets to intelligently manage the switch between feature delivery and addressing tech debt. Or X% time for engineers to focus on tech debt. These strategies are well proven but rely on the discipline to stick to the model. In my experience it’s often too tempting for organisations to ‘steal time’ back rather than ‘stop the line’ and these strategies get watered down.

This is where AI tooling can be a game changer. AI tooling used with these strategies leveraging existing knowledge can start to be brought to bear on tech debt. It can be used to minimise downsides and avoid the time stealing problem. Because:

• Using existing knowledge and organisational context, AI can be used to quantify the value
• AI can be automated to react to error budgets
• AI can work on the remediation tasks 100% of the time. The work can be defined and specified within the context of the organisation and low risk work farmed off and addressed immediately.

High value revenue driving new features that can’t be easily derived from existing knowledge can be done by skilled engineers who will no longer need to attend to as much of the tech debt.

Having that existing knowledge and legacy provides a strong context to direct and inform AI how an organisation functions. It can be used to form the guardrails for effective and on-going transformation. Legacy is an asset. It allows organisations to maximise the value of an existing business. It can super power and automate proven strategies for tech debt remediation.

Don’t Fear the Rewrite

In any sufficiently long lived enterprise there is always at least one thorny system that could do with a complete rewrite for one reason or another. We are reluctant to, or fear, doing this. Often for good reasons. Like the strategies for weaving tech debt remediation into other value streams, strategies such as the strangler pattern have been defined for modernisation of legacy which avoid the trap of a ‘big bang’ rewrite.

There are several ways to apply AI tooling here:

• It lowers the bar on a ‘big bang’ rewrite of some systems. Mitigating some of the key risks of doing it. Making it a more tenable approach in some cases
• It can be used to help with the analysis required to find the seams required to break a legacy system into components and sequence them for modernisation by a strangler approach. Accelerating it
• It can help to continue to deliver business requirements against the existing system while engineers are able to work in parallel writing new

This is all to say that the smart application of AI technologies really can lower the gravitational drag that legacy and tech debt put on an organisation. Enterprises can make a real leap forward today where they may have been stuck yesterday. Again, legacy can be used as an asset to provide the organisational context required for effective modernisation.

Knowledge + Experience + AI = Win

There was a cute phrase going around a while back that said “AI won’t take your job, but someone that knows how to use it will”, while I think this was partially said defensively rather than optimistically. There is some truth in it. What does “using AI” mean?

You will have noticed the reference to several existing strategies and approaches to managing change within an enterprise. These are extremely effective and valid approaches. Those organisations that can apply AI to proven ways of working and leverage existing knowledge will accelerate the fastest.

Experience and understanding of these approaches will allow engineers to be effective with AI rather than replaced by it.

Applying AI like a nuclear bomb rather than a surgeon’s scalpel in an attempt to achieve short term gain will likely end in long term disaster. Unfortunately due to the hype and promise of immediate value which has driven a high level of investment in AI by some organisations, will result in pressure to show value quickly. Making the cost avoidance argument a hard sell.

In Summary

AI tooling will work well within established businesses where organisational context can be gained from existing legacy and an existing body of knowledge.

If AI tooling can also:

• Be effectively deployed to accelerate remediation work that has traditionally slowed larger enterprises down
• Address areas where it has been tough to apply resource to without stealing from other value streams
• Used to quantify and execute on modernisation efforts
• Deployed in the context of existing approaches and ways of working likely to already be familiar and established

Then:

• Highly skilled and motivated engineers can be utilised more effectively to power direct revenue generating change
• Enterprises can accelerate in areas they traditionally struggle
• Existing strategies already in regular use can be super powered

Finding ways to employ AI to strip out the work that is both unpopular for engineers and less favourable to the business is win/win. The business gets to see more value and engineers are able to do more rewarding and innovative work.

Looking forward to both seeing how this plays out in 2026 and being part of the story.