01732 759725 25 magazine sources and watch for signs of manipulated content or fabricated references. Embrace the idea that you will ‘fail your way to success’. The pace of change in AI is so rapid that if your strategy is to wait for perfect information you will quickly fall behind. Begin with low‑risk, high‑frequency tasks: meeting summaries, email and document drafting, light analysis with human cross‑checks. Run short pilots and keep what works, drop what doesn’t. The goal should be momentum, learning and measurable outcomes, not a grand plan that goes nowhere. Borrow patterns from larger organisations, then simplify. Enterprises will often publicise what works well. Your CoE can translate these into a right‑sized set of guardrails, avoiding reinvention while minimising overhead. Finally, understand how disruption varies by sector and by task. Legal, marketing, customer service and elements of finance may see faster automation than asset‑heavy manufacturing for example. Map your work by tasks and assess the predictability of each task, the level of document/ data involved, its frequency, risk levels and visibility to customers. This will give you a heat map showing where AI can create value quickly and where extra controls are essential. A natural adoption journey Given these guidelines, what might a sustainable journey to achieving maximum value from AI look like for an SME? Its starting point should be where friction is lowest, i.e. within the tools people already use, as this will generate immediate payoff and generate curiosity to go further. Begin with AI assistants for everyday tasks, like meeting transcription and summaries. Then, as confidence grows, move from isolated tasks to small workflows: link an AI‑assisted summary to a follow‑up plan, for example, or pair a proposal draft with a review checklist. At this point, patterns will start tools that accelerate analysis etc., because they deliver value quickly and build confidence. However, the real inflection point came when we zoomed out and started to think how work could flow across the organisation if everyone had AI assistance. This meant asking different questions. If support can predict incident patterns, how should services be designed? If sales can automatically triage and enrich enquiries, how does the funnel change? If data analysis is faster and richer, where can you shift time towards discovery, customer conversations and creative problem‑solving? Tips for SMEs considering AI SMEs often tell us they want to move forward with AI but feel overwhelmed at the prospect. Based on what we’ve learned and what we’ve seen work in our business, we offer the following advice: Start by putting simple governance in place. Clarify which tools are approved, which data they can touch and which tasks need human oversight before anything leaves the building. This makes early experimentation productive rather than risky. Create a lightweight AI Centre of Excellence. In many SMEs, two or three people are enough: one with security or IT responsibility, one with data or analytics experience, and one from a business function such as operations or finance. Give them a clear remit to monitor developments, publish standards, coach teams, track outcomes and keep adoption aligned with business goals. This will empower employees to drive AI innovation supported by a dedicated team and can help guard against Shadow AI. Be vigilant. Bear in mind that tools differ in their maturity, and even the best can be wrong. Always be on the look-out for ‘hallucinations’, where the AI model confidently generates incorrect or misleading information. Teach people to verify claims, check process thousands of customer support tickets every day and troubleshoot hundreds of different software and hardware solutions. Trend analysis means we can now see – across the whole company – which technologies generate recurrent issues, enabling us to guide proactive support, identify automation opportunities, steer future product recommendations and surface potential new solutions that we can take to market ourselves. AI has also reshaped our workforce plans. Initially, we anticipated hiring more data scientists, but in practice, better tooling and embedded capabilities have reduced the need for heavy bespoke modelling. As a result, we are now focusing on AI‑literate analysts, solution architects and product owners – roles that combine domain knowledge with the ability to wield AI responsibly and effectively. This shift has boosted productivity, allowing staff to spend less time on mundane tasks and more on tasks that require judgement and on developing relationships. Governance must evolve as fast as the technology Traditional IT policies can remain unchanged for years, but not AI policies. New features arrive relentlessly, and the way models behave changes all the time. We have learned to update our AI policy frequently and to treat it as a living document. The substance of that policy is straightforward: data access should follow least‑privilege principles; sensitive outputs require human review; confidential information must stay within approved environments; logs should provide auditability for significant automated actions; and adoption should be measured alongside outcomes so we can course‑correct. From small wins to reimagining the business At Phoenix Software, we deliberately started with many ‘small things’, e.g. agents that summarise emails and meeting notes, assistants that draft content, AI continued...
RkJQdWJsaXNoZXIy NDUxNDM=