Practical tips for successful AI adoption in the mid-market
Alper Gunaydin, CTO at Klarus, explores the barriers holding mid-market companies back from scaling AI projects and offers five tips for successful adoption.
The conversation around AI has largely moved beyond whether businesses should adopt it. The bigger challenge now is turning AI usage and experimentation into meaningful, scalable business outcomes.
Klarus’ recent research into the mid-market found that nearly three quarters of mid-market companies in the UK and Ireland have already moved into deployment, with 31% fully deploying AI across multiple business functions and 42% partially deploying it in specific areas.
However, 90% of those that have explored AI in some capacity still have some initiatives in the early stages or have experienced stalled pilots. What’s clear is that many are finding it difficult to achieve their ambitions and scale efficiently.
The barriers to moving beyond pilot
The research found that lack of AI expertise and concerns around governance, ethics, security and privacy were the joint top reasons pilots stalled, each cited by 48% of respondents that had explored AI.
Almost half, 45%, also said AI paralysis, not knowing where to start or facing too many options, frequently or always slows progress. Data and technology foundations were a further barrier, with 83% of companies that have piloted or deployed AI experiencing poor data quality, and 69% saying this was preventing or delaying AI projects.
Based on the research findings, we’ve put together five practical tips to help mid-market companies overcome the challenges and move from pilot to deployment with greater confidence.
Five tips for AI adoption
- Start with the business problem, not the technology
It’s important to establish a clear business case before investing in an AI solution. That means defining the problem AI needs to solve, the process it will improve and the commercial outcome it should deliver before selecting tools.
An important measure of success for any technology transformation project is return on investment (ROI) and the findings from our research are encouraging. A notable 85% of AI projects have successfully met or exceeded their ROI expectations and only 11% underperformed, with 3% having no way to measure.
To replicate and continue this success, set clear ROI expectations from the outset. This includes both direct benefits such as cost savings or revenue uplift and indirect benefits such as productivity, quality, risk reduction and capacity release. The calculation horizon should then be determined carefully, considering the nature of the technology investments.
Breaking out of stagnation can also partly come down to mindset. Remember that you can also learn by implementing. Ruminating over where to start can be a barrier in itself, so treat early projects as opportunities to test, learn and refine rather than waiting for the perfect AI strategy.
- Prioritise practical use cases
Only 10% of companies that have explored AI have successfully scaled all of their initiatives beyond the pilot stage.
To avoid getting stuck in pilot mode, focus on areas where AI can deliver a measurable benefit rather than trying to implement it everywhere. So, instead of chasing broad transformation from the outset, try prioritising targeted use cases that can deliver measurable outcomes and that can easily be integrated into existing workflows.
The goal should be to move away from one-off pilots and trial models that can be repeated at scale across your organisation.
- Prioritise data quality before scaling
Klarus found that 83% of companies that have piloted or deployed AI have experienced poor data quality, and 69% said it was preventing or delaying AI pilots or deployments.
Poor data quality is one of the biggest barriers to AI deployment, so treat data readiness as a core part of your AI strategy rather than a technical after thought. Improving data accuracy, ownership and accessibility will pay dividends, helping accelerate pilots into production and ensuring projects stand the greatest chance of success.
That said, data quality shouldn’t stop you from starting your AI deployments. This work can be done in parallel, provided it is considered and integrated from the outset.
- Embed governance, security and ethics early
Establishing strong guardrails for AI projects is imperative and it’s where a lot of companies face challenges. In our research, more than half of respondents, 59%, had yet to establish a comprehensive AI governance framework, while governance, ethics, security and privacy concerns were cited by 48% as a primary reason pilots stalled.
When mapping out AI projects, design guardrails from the outset rather than wait until a pilot is up and running. Having clear policies around privacy, security, accountability, human oversight and acceptable use will all help reduce risk. They will also help increase stakeholder confidence in projects and ensure there is strong sponsorship, which is important to preventing projects from stalling.
- Build a blended team model
Lack of AI expertise is the joint top reason for projects failing to move beyond pilot stage. Building internal expertise is also an area that 39% of respondents in our survey are prioritising over the next 12 months.
The best approach is to combine internal ownership with specialist external expertise. Internal teams understand the business context and it’s important to build capabilities and give teams the skills and confidence to use AI effectively and responsibly.
It’s also important to know when to bring in specialist expertise. That means recognising where external support can help you move faster and avoid costly mistakes. External specialists can often help navigate the AI vendor landscape, validate use cases, design robust delivery models and close skills gaps more quickly.
Many mid-market companies are opting for this hybrid approach, with nearly half of respondents (44%) using a combination of internal teams with specialist external support.
A roadmap to deliver value
Mid-market companies have moved beyond AI experimentation and have their sights set on further scaling their deployments. Success, however, hinges heavily on having the right execution capabilities.
Those that can link AI to clear business outcomes, put in place strong data foundations, embed governance from the outset and blend internal expertise with specialist support, will stand the best chance of turning experimentation into measurable business value.