Why AI isn’t going to cure all SMEs’ productivity problems

From integrating ChatGPT into internal operations to catastrophizing how the technology could overtake hundreds of jobs, artificial intelligence has found its way onto the agenda of company executives.

For large corporations like Google or Microsoft, developing AI products and using large language models is a no-brainer. With their resource coffers and central role in the tech world, they’re de facto expected to lead the charge in emerging technologies.

For small businesses, the answer isn’t as obvious. Although 55% of SME owners understand what AI can do, the majority still feels nervous about integrating it into their business strategy..

Beyond the haziness of what AI really is and how SMEs should think about adopting it lie more important questions – do SMEs even need to be developing their own native AI systems? Is AI being put on a pedestal as the be-all-end-all to solving a company’s operational issues?

After speaking with EMERGEiQ, a data science solution company that designs AI APIs for small businesses, the short answer is no – SMEs don’t need to go from zero to 100 when it comes to AI. Most times, it’s not even the tool they need to bolster enhanced productivity.

AI is great but it’s not for everyone

Around one in six UK organisations, totaling 432,000, have embraced at least one AI technology according to government research.

If we zoom into how this is distributed across company size, 68% of large companies, 33% of medium-sized companies, and 15% of small companies have incorporated at least one AI technology.

According to Startups’s research, funding for AI startups has increased by 66% over the last three years. Investors are interested in AI, which is why startups who work in this space are seeing their account books benefit from the hype.

Despite all the enthusiasm, there is a hefty deja vu feeling that serves as a warning sign to SMEs who are still unsure about whether AI is for them.

Turning back the clock to the late 1990s, the world wide web became a differentiator factor for investors. As the hype around digitising business and building it a piece of internet real estate grew, so did the dot-com bubble. Until it burst, leaving public stocks of internet companies in tatters.

Drawing on these lessons, Mizan Rahman, Co-Founder and CTO of EMERGEiQ, recommends that SMEs approach AI with purpose.

“It’s still not particularly cheap to build something proprietary and do something yourself,” he explains.

“Unless you genuinely want to, if there’s a need for your organisation to have a reasonable and impactful transformation to make significant cost savings and potentially ROI, then you don’t really need sophisticated AI use.”

Mizan shares that when EMERGEiQ has been approached by potential clients to develop an AI solution to cure their productivity ills, most of the times, AI isn’t always the answer.

“We’ve met companies that have legitimate user cases but they don’t have the data set yet to embark on an AI journey – wait a few years, build your data set, clean up the data – you need it structured and cleaned out,” he warns.

The devil is in the data

The difference between a genuinely useful AI system and a poor one that barely feeds back coherent results is the data upon which it is built. Cultivating, sorting out, and cleaning up the data takes time, and most importantly, resources.

The barriers of entry into AI are still high. Not only is it expensive, but it’s difficult to get access to the quantities of data necessary to build an advanced large language model.

To put it into perspective, the development of ChatGPT cost OpenAI around $4.6m (£3.7m) and that was only to develop the first version of the chatbot.

Although the release of open-source models like Llama 2 by Meta is an indication that eventually the barriers to entry could be lowered, the keys to the data kingdom remain in the hands of those that have the resources to pay for it. Think Google, Meta, Microsoft, or national governments.

These high costs mean that a company’s definition of digital transformation needn’t include advanced artificial intelligence.

“Relative to the size of the organisation, they should be thinking about what their future state should be,” reveals Rahman.

“If you’re a very small organisation and operationally you’re doing okay, you’re making ends meet and you’re a traditional type of organisation, you could be thinking about using technology for customer engagement, loyalty, and other kinds of transformational requirements,” he adds.

Most importantly, SMEs have access to tools that themselves integrate some form of AI, saving them the expense and time of developing their own native API.

For instance, Salesforce products now feature Einstein AI in their CRM solutions. Canva can seamlessly create graphics for social media by prompting its artificial intelligence function. Businesses can even create AI-generated websites with website builders like Wix.

Referring to these more cost-effective ways of using AI, Rahman believes “SMEs will still say that they’re using AI and they will integrate these existing platforms and these existing services into their workflows which is admirable, because they are trying to improve things.”

“But shifting over to a regulated, more controversial kind of potential in-house built stuff, that’s still too expensive and you have to be serious about integrating sophisticated AI,” he points out.

Making AI count

With the recent conclusion of the AI Safety Summit and the signing of the Bletchley Declaration on November 2, it’s clear that even governments are still trying to wrap their heads around the impacts the technology will have.

From setting regulations in stone to increasing investment in AI, there’s still a lot of pending question marks and milestones to achieve.

Rather than trying to preempt and do guesswork, SMEs are wise to approach the technology carefully and question how necessary it is to invest in an in-house solution.

As Rahman emphasises, the market will do its thing and eventually lower the barriers of entry to AI as the technology evolves and becomes easier to tame.