Overcoming the challenges of AI adoption in SMEs

Sandy Kahrod at Six Degrees outlines the key challenges organisations face when integrating AI technologies into their existing workflows and what organisations can do to ensure successful AI integration

 

At the same time as AI is permeating our personal lives, its uptake across businesses is increasing rapidly. However, while the technology is readily available with solutions like Copilot and ChatGPT, integrating AI into work practices is not as straightforward as it first appears.

 

Organisations, particularly SMEs, are often unaware or unprepared for the challenges an AI deployment can bring. Whether that’s appreciating the upfront costs, incorporating the right security, or deciding which areas to prioritise, it’s important to understand all the implications and have a fully-fledged plan in place.

 

Advance preparation and realistic budgeting will also assure smaller organisations that AI projects don’t necessarily require multi-million pound budgets to be successful. SMEs can take advantage of the capabilities of AI at an affordable cost if they follow some fundamental guidelines, and don’t go for a big bang approach.

 

Where to start

Start with a well-defined project, where the workflows involved are clearly documented and outcomes from deploying AI are measurable. Avoid choosing areas where it is difficult to assess current levels of efficiency or productivity, as it may not be apparent whether AI has made improvements or if the original process itself was flawed. Instead, focus on a specific activity where success can be tracked.

 

For example, an online retailer might lose sales during promotions if they cannot predict stock requirements accurately and alter deals and discounts to reflect this dynamically. AI can manage these changes on the fly, calculating stock levels and optimising offers in line with availability of products. Like-for-like comparisons with previous campaigns would demonstrate the difference AI made to sales volumes and margins.

 

AI subject matter expert

Ideally, take on a subject matter expert when looking at the potential for rolling out an AI tool like Copilot company-wide.  This will help marry up the capabilities of the AI tool with the issues it could solve on a departmental and function-by-function basis. Otherwise there can be unrealistic expectations that the tool will solve everything! Instead, recognise each area has its own unique set of data and processes. 

 

Additional resources are likely to be needed to document findings and translate them into the requirements for the AI project. This ensures models are built accurately to give appropriate responses and outcomes. 

 

Alternatively, use external resources and third parties who can provide deeper expertise and experience. 

 

Data privacy and accuracy

Be mindful that GenAI tools work on huge amounts of data and, while powerful, they still don’t always behave as expected. The rogue chatbot at Air Canada is just one example of how things can go haywire and cause embarrassment - so much so that, in this instance, the misinformation it provided to customers led to a lawsuit. Therefore, caution should continuously be exercised when putting data, old or new, into AI, whether that’s customer, financial, employee, or IP. Remember, any data imported could be inadvertently exposed. After all, no one wants to be responsible for the AI tool that ruins a company’s reputation! 

 

Data may also contain biases towards gender, race, age, and other factors that could distort results.  Identify gaps or concerns in advance. Build in governance processes, testing, and feedback loops to ensure data privacy and accuracy are paramount, and expect errors as the first iterations will not be perfect. Incorporate compliance obligations for privacy and security regulations, including GDPR and NIST 2. Put in place contingency plans for when things go wrong. Learn from mistakes!

 

AI integration

It’s hard not to get caught up in the enthusiasm for AI with vendors offering free POCs and encouraging organisations to play with the tech. However, it’s important not to lose sight of the cost of integration and possible changes to the underlying infrastructure.

 

This includes data preparation and security. It’s much more costly and complex to clean and secure data properly as an afterthought. Make sure the business case for each AI project is fully-rounded and covers all aspects of data preparation, system integration, and security protocols - then there should be fewer nasty surprises along the way.

 

Change management

Any significant changes in work practices are unsettling, and there’s no doubt that AI will have a huge impact on jobs. But misinformation has caused unfounded concerns about swingeing job losses. By taking the time to put comprehensive change management programs in place, organisations can reassure employees about the introduction of AI. 

 

From the outset, showing staff how their jobs can evolve rather than be replaced sets a positive foundation for what’s to come. Also, support people through the change period by setting up regular update sessions, having internal champions to help and listen to feedback, and encouraging ongoing engagement with weekly tips and guidance. It takes time to change habits and preconceptions, but with consistent positive messaging, most can be converted.

 

A fast fix?

AI is definitely not a fast fix. But its vast processing power can be applied successfully to many processes and problems across all sizes of organisation with the right preparation and planning. For SMEs, cloud-based tools and open-source options with user-friendly interfaces are making these capabilities more attainable and cost-effective than ever before.

 

Without the need for bottomless AI budgets, SMEs can develop tailored solutions, increase efficiency, and grow, proving that AI’s benefits are not exclusively for the enterprise world.

 

---

 

Sandy Kahrod is Modern Work Product Manager at Six Degrees

 

Main image courtesy of iStockPhoto.com and Tzido