By Saravana Kumar
As AI spending surges, CEOs face pressure to prove ROI. This piece by Saravana Kumar, founder and CEO of Document360, looks at how to avoid hype-driven decisions, and build sustainable long-term value.
The last few years have seen AI move from novelty to board-level mandate, but when you step back you can see that the “goldrush” phase is creating a familiar pattern: overinvestment, inflated expectations, and rushed pivots. However, I believe that companies with capital discipline are uniquely positioned to harness AI responsibly, because they must prove ROI, protect runway, and build for long-term customers, not short-term cycles.
The goldrush is real, but so is the hangover
If you are lucky enough to lead a software business today, you can feel and see it: AI is no longer a mere side project. Our product roadmaps, hiring plans, and so often budgets are being rewritten around it. At the same time, markets are signalling that “AI as a story” is not enough, with Reuters recently reporting that a software-stock selloff has disrupted dealmaking, with uncertainty about how AI reshapes business models adding to valuation volatility.
Adoption is accelerating in parallel. According to Gartner, by 2026, more than 80% of enterprises will have used generative AI APIs or deployed GenAI-enabled applications, which is up from less than 5% in 2023. What does that mean for CEOs? Well, as well as opportunity, it presents a combination of risks: widespread experimentation alongside rising pressure to show outcomes.
Why capital discipline changes the AI conversation
I’m a technical founder who moved from Mettupalayam near Coimbatore, India, to London and built Kovai.co across the UK, USA and India. Today, we employ roughly 280people, and Document360 has grown to $10M+ in annual recurring revenue while serving more than 1,500 organisations. That background shapes how I think about AI: every investment has to earn its keep. When you build on customer revenue rather than investor runway, you can’t hide behind narratives for long as every pound spent on AI competes with investment in reliability, security, and customer success.
Capital discipline also keeps you from mistaking motion for progress. The AI wave risks repeating a mistake many tech companies made in the free-capital era: scaling spend before proving economics. Disciplined businesses habitually behave in a different manner. They prioritise what you could describe as the “painkiller” (solving real, urgent problems) use cases over “vitamin” features (the nice to haves). These types of businesses also test small and only then scale on what actually works, and keep optionality so they can invest through cycles rather than pivot with them.
Pragmatic AI in enterprise SaaS
One lesson I have learnt from a career in enterprise SaaS is that this type of software teaches us a sense of humility. In our world, mature products are rarely “replaced”, rather they are refined. AI is a powerful accelerator, but it cannot substitute the unglamorous work of enterprise-grade engineering: permissions, audit trails, availability, and support. Gartner has also warned that hallucinations and inaccuracy can limit the impact of GenAI-enabled applications if they lack guardrails.
The best AI deployments I’ve seen follow a simple principle: use AI to compress time-to-value. In knowledge-heavy workflows, that means helping teams draft, find, summarise, and maintain information faster, without breaking trust. In our work building Document360, we’ve described how AI can help technical writers produce content in hours rather than days, and how dozens of AI-driven features can be shipped as incremental improvements rather than a risky “rip-and-replace” overhaul.
AI is also changing monetisation. Rather than the standard pure seat-based pricing, AI is now pushing vendors towards more creative credit or usage-based models, so customers pay for outputs (such as work completed faster) rather than simple logins. In turn, buyers pull finance and procurement into AI conversations earlier, asking for baselines and proof, not just demos. If you can’t measure and prove productivity gains, you will struggle to price AI credibly.
Evidence that “boring” systems drive real ROI
One reason AI initiatives disappoint is that companies underestimate the infrastructure beneath them and so it is asked to perform without a stage. Knowledge management, documentation, and internal workflows – these are not the most glamorous of pursuits, but what they do is provide one of the highest-leverage inputs for AI. If you don’t adhere to this discipline, your information can be scattered or outdated, and AI only will amplify, at speed, that confusion.
Consider Triton Digital, which described the weary operational drag of maintaining self-hosted documentation: even small text changes required recompiling and re-uploading, sometimes multiple times across 24 hours – an administrative Groundhog Day. After moving to a cloud-based documentation site, they said updates could be made in minutes and reported a reduction in support calls and time spent handling those calls. AI alone did not create that impact; a stronger knowledge system did.
A CEO checklist for surviving the AI goldrush
Below is a framework for those who want to win in AI without burning runway.
| Decision area | Capital-discipline test |
| Use case selection | Will this remove a measurable bottleneck for customers or internal teams within 90 days? |
| Data & knowledge | Do we have reliable, permissioned knowledge to ground AI outputs and reduce hallucinations? |
| Product maturity | Are we improving a trusted workflow, or adding AI to compensate for gaps in the core product? |
| Risk & compliance | Can we explain outputs, manage sensitive data, and meet emerging regulatory expectations? |
| Business model | Can we price AI based on value delivered (time saved, tickets deflected, risk reduced)? |
This checklist is not about slowing down or hesitating. It is about moving with velocity but ensuring you do so in the right direction. It is about protecting the runway for work that is open to scrutiny and genuinely compounds value. If a project cannot survive these questions and justify its very need for existence, it probably needs to be smaller, or perhaps even stopped.
Closing thoughts
The AI goldrush has and will continue to produce winners, certainly, but not necessarily the loudest or the extravagantly financed. It will reward companies that combine controlled ambition with decent stewardship: shipping useful improvements, proving ROI, and protecting runway. Capital discipline, in this light, is not a restraint to growth; it is the advantage that lets you keep building when others have to stop.
Born in Coimbatore, India,
