Despite the hype surrounding the promise of AI, the true cost-to-value ratio of this transformative technology remains elusive. The challenge lies in the multifaceted nature of AI’s value proposition: Whether it’s from a product, compliance, talent, infrastructure, or even environmental standpoint, the true costs of AI at this point remain shrouded in uncertainty.
As organisations pinpoint the use cases that deliver measurable, tangible value – all while keeping an eye on the costs associated with those endeavors – how can they best navigate a path forward without making too many missteps along the way?
Don’t be shortsighted about security
Looking back to the early days of the internet explosion provides some useful lessons for this new era, particularly when it comes to teasing out cost and value around AI.
In the early internet era, many teams prioritised growth over guardrails. Simply put, cybersecurity was not a priority. Money saved by not addressing the security aspects of the new technology allowed for quick growth of functions and use cases – but just as quickly, this created huge costs in reputation and lack of trust in the whole paradigm. Ultimately, of course, this cost companies much more on the backend when they had to retrofit security controls or add in breach mitigation functionality.
The lesson for the AI era? It is critical that elements like security and privacy are embedded in the development of AI as early in the process as possible to reduce re-work, retrofitting, or not providing the type of risk identification and remediation features that are necessary to avoid regulatory fines or breaches. Sprinting ahead with AI without taking security and privacy into consideration is simply shortsighted.
Give governance room to breathe
When it comes to AI, there needs to be governance from day one. A governance charter is a good starting point, but AI is a fast-evolving domain. Compounding the challenge, the surrounding regulatory landscape is in constant flux, with new laws and compliance mandates emerging seemingly daily.
The best advice for organisations here is to not chase perfection when it comes to their governance charter. What’s most important is that there actually is a governance charter in place to guide their AI efforts and – more importantly – that it is revisited on a regular cadence. There should be a general understanding that the governance model is a work in progress and that there will be continuous, ongoing improvements. Don’t let perfect be the enemy of good.
Build resources without breaking the bank
In another echo of those heady internet days – when software engineers could often name their price – the cost of knowledgeable AI resources is spiking. There are signs in the market of big players over-spending for talent, likely calculating that early success will provide long term advantage and create higher barriers to entry for other players.
For most organisations, this level of over-spending is not an option. Instead, focus on growing skills in-house. A combination of supported and governed AI experimentation, ongoing investment in education, and collaboration between business and technical resource teams will develop the skills and targeted use cases that can accelerate value realisation.
As resources become skilled, they will be susceptible to market demand, so the supporting compensation for these skills need to be reviewed and adjusted frequently – ensuring that costs are optimised to the value provided.
Focus on clean data and a clear risk assessment
The primary “fuel” of AI is curated data. Invest early in understanding what data your models actually need, then build automated pipelines for ingestion, cleansing, de-duplication, and gap identification. Organisations should treat data curation as a continuous process, not a one-time cleanup. Fortunately, the upside of this effort extends beyond AI – disciplined data governance benefits multiple aspects of the business.
Also: it’s essential to understand the regulatory risk around your usage of AI. When using third-party AI embedded in operational systems, organisations must understand their vendors’ philosophy, ethics, and compliance posture. Classifying AI applications by regulatory risk category is also critical, especially for highrisk uses that require defined points of human oversight.
For AI built inhouse, start with crystalclear expectations: what problem you’re solving, how outcomes will change, and whether the goal is efficiency, insight, advanced automation, or some other business objective. Align data quality, processes, and model design to those objectives from the outset, so the solution is built on a foundation that supports the intended business impact – making it best positioned to drive value.
Be aware of “the hidden costs”
The “hidden costs” of AI implementation include infrastructure, compute, integration, energy, staffing, governance, and ongoing maintenance, amongst other factors. These variable costs can vary greatly depending on the scope of the use case.
Along with the energy costs associated with AI, there is already a growing public awareness of the effect that AI power consumption is having on the environment (to say nothing of the costs to everyday utility consumers who are seeing higher bills due to the increased overall demand).
It’s not hard to imagine an upcoming swing of the pendulum towards “responsible corporate citizenship”, where it becomes a positive PR selling point for companies to measure the impact of their AI usage and have a strategy to reduce that impact so that any benefits derived from AI aren’t unintentionally dragged down by a reputational “cost.”
Think long term
Unpacking AI’s costtovalue ratio demands more than chasing the next breakthrough – it requires mindful choices at multiple levels. Security built in from day one, governance that adapts as fast as the technology, sustainable talent strategies, clean data, and clear risk alignment all feed the equation. The organisations that thrive will be the ones that resist the temptation to chase quick wins and instead treat AI as a longterm strategic bet – one that can deliver enduring value long after the hype fades.
