How to implement AI as a mid-sized company: 5 practical steps to success

48% of businesses are currently experimenting with AI technologies – as revealed in Microsoft’s recent report, ‘Accelerating competitive advantage with AI’. That means, for the majority of organisations, particularly mid-sized businesses, the need to implement AI simply isn’t on their radar.

In truth, it’s likely that companies like this didn’t realise that the internet was a ‘must do’ at the turn of the millennium either. That’s why it’s essential to help these sorts of businesses understand not just the importance of technologies like this, but how they can integrate them across the organisation.


Being held back

In my experience, there are three main reasons why AI opportunities can be difficult to see.

  • Lack of knowledge about the technology – which leaves forward-thinking companies to gain the competitive edge
  • Case studies aren’t relevant – in most instances, case studies are about, and aimed at, global brands with a near-infinite supply of money
  • Tech projects, and particularly AI projects, have notoriously high failure rates due to lack of experience, complexity of the project, and unrealistic expectations – no-one wants to face spiralling costs, white elephants, and business disruption

The trick to successful implementation of AI is to refuse to over-commit at the early stages – this, instead, should be the exploratory stage where a business discovers AI’s impact, and how it should be deployed.

Once the impacts are fully understood it’s time to develop a roadmap, incorporating each part of the business that will use or be affected by the technology. This leaves room for smaller initiatives to be drawn up and integrated into the ‘big picture’ programme.

Oversight is also critical to success. The team in charge of executing the AI master plan needs to engage with the rest of the business. That way, they can evaluate success (or potential failure), serve additional knowledge, and give the leadership team necessary feedback.

With this in mind, I believe there are five core steps to making AI integration a success.


Step 1 – Teach the benefits of AI

Gaining internal buy-in at the earliest possible stage is vital – this starts with the leadership team. AI is not just another IT project, after all. You should then look at which areas of the business will also be benefiting from the technology and get them involved in the process. It won’t be long until AI is essential for many operations across a business. Don’t get left behind.

Appoint someone to own your AI project. This should, ideally, be someone with strong leadership skills, as they’ll be heading up a cross-functional team and informing company leadership of its progress. Your ‘project owner’ should also be considering the skillsets and expertise needed to bring the project to completion, whether internally or with your partners.

It’s their job to drive through success. Everything must be measured and kept in control. It’s the only way to dodge those high failure rates.


Step 2 – Build a team or partner up

Depending on the size of your company, this is the ideal moment to construct a multi-disciplinary team. Keep this to around three people. It’s likely you’ll need to gain a fundamental understanding of AI, the opportunities, and the challenges. This overview helps guide you as you build out the required capabilities of AI and your proposed solutions.

Seek out partners with experience of implementing data and AI projects for similarly sized businesses. These partners will have a good understanding on the challenges you may face, and how to circumvent them. As with any IT project, but especially AI implementation, the sooner you get them on board, the easier it will be.


Step 3 – Identify the right problems

Why do you want to implement AI and what do you plan to do with it? Begin by building a solid business case, focusing on key challenges, and how AI can help overcome them.

If AI can’t solve these problems, look at where your organisation could become more efficient through automation, instead. For many smaller companies, this will prove an excellent stepping stone to future AI adoption.

Indeed, given failure is linked to a project’s overcomplexity, it’s worth considering whether your first AI project could be used to automate a simple process. You don’t have to re-build your business from the ground up with an AI foundation. It’s all about giving you time back to focus on more critical tasks.

I recently worked alongside suit e-tailer The Drop. During that time, we identified three areas where AI could benefit business process, while improving the experience for customers.

  • Finding mistakes in customer measurements
  • Measuring without a tape measure
  • Create a 3D image of a customer using two photos

For The Drop, AI helped them streamline the supply chain. This resulted in faster deliveries and fewer returns, since measurements were accurate and mistakes were found before production.


Step 4 – Get your data ready in advance

AI is only ever as good as the data it has. So, once you’ve identified how AI will assist you, you must then look at what data you have or need to ensure it works properly. Your data should be clean, organised, and easily accessible for the technology.

A word of warning, though. It’s possible your IT team may push back on this request – but the data is owned by the entire business, not a select department. As long as that data is secure, flexible access is a must to capitalise on the value it holds.

The Wild Me project is a good example of using data – specifically, images – for the greater good. Anyone is free to upload animal images to the Wildbook Cloud, where they can be catalogued and tracked. That crowd-sourced data then helps scientists make informed decisions over conservation efforts.


Step 5 – Activate with AI

With the other four steps complete, the final step is simple. With your business identifying a business challenge, how AI can solve it, and the data it needs to do so, you can start rolling out your AI initiative.

Start with a proof-of-concept. It helps you achieve your stated scope and scale of the project. Next, create a model; build and implement AI. Test it. Ensure it’s delivering what you need. You’ll be obtaining results and value in no time – whether that’s increasing process efficiency, analysing data, or making customer experiences truly personal.


And I say all this as someone who has been on the journey you’re preparing for – at the start, we used Microsoft’s tools and products, like the pre-built Cognitive Services, and the cloud capabilities of Wirehive to build a chatbot. With the initial groundwork done, and the experience gained, we’ve been able to work on even more complex projects, fuelled by increasing amounts of data and machine learning models. If we can do it, so can you. Good luck on your journey.

[msce_cta layout=”image_center” align=”center” linktype=”blue” imageurl=”” linkurl=”” linkscreenreadertext=”Link to download Microsoft’s AI report” linktext=”Download the full AI report” imageid=”17871″ ][/msce_cta]

Find out more

Become a Microsoft Partner

About the author

Headshot of Dean CorneyDean has owned and managed agencies for over 16 years. In this time, he’s worked on a range of technology projects for global companies. Now, in his role as COO at a specialist health and beauty agency, The Pull Agency, he creates multi-disciplined teams that help bring the worlds of brand and technology together. Although Dean has a broad, journalistic knowledge of emerging technology, he’s most passionate about solutions that improve customer experience, applying old school marketing know-how and creative thinking with the latest technologies. He has a passion for road cycle racing, having represented Great Britain at the Under 18’s. Dean also enjoys a good Rioja and listening to Radiohead at maximum volumes.