How to calculate and maximise the ROI for AI deployment

At MakoLab, we have a long history of helping organisations deploy technologies that not only work but, primarily, deliver measurable financial outcomes. Similarly, AI simply can’t be treated as just another cost on the list. It’s a strategic driver of growth with potential that both should be calculated and is well worth calculating.

Return on Investment (ROI) is an indicator that answers the business world’s most straightforward and vital question; how much did we earn in relation to the amount we spent?
Simple? Only outwardly. When it comes to artificial intelligence, calculating ROI can be a challenge. This happens because the benefits of the technology are multidimensional and often spread out across time, while the costs are considerably greater than the price of just purchasing some software.

The price of a licence or monthly subscription for an API is merely the tip of the iceberg. To obtain a reliable calculation of your ROI, you need to know what the total cost of ownership (TCO) is.

At MakoLab, we always take four key areas into consideration.

·     The technology: This isn’t just a fee for software. It’s also includes enormous computing power, in other words, the costs of either a public cloud like Azure or AWS or your own on-premise server, the costs of integrating with your existing ERP and CRM systems and the costs of data transfer fees.

·     Human resources: Someone has to implement the technology and supervise it. This means the salaries, or time, of data scientists, MLOps engineers, project managers and the business analysts who ‘translate’ the company into the language of algorithms. The crucial costs of training the rest of the team also need to be included here.

·     Data: Data storage infrastructure, like data lakes, for instance, is one cost. However, the hours spent on extract, transform and load (ETL) processes, which is to say, cleaning, standardising and labelling data, is a real, often hidden cost. Remember: AI is only as smart as the data it learns from;

·     Maintenance: An AI model isn’t a building that’s constructed and then just stays where it was put. It’s more like a garden. It needs to be watered, by way of providing new data. It has to be pruned, which is the ongoing optimisation process. It must also be protected from weeds, in other words, monitored for ‘model drift’, a situation where the model no longer fits a changing reality.

Now we know what the costs are, it’s time to seek out the profits. The effects of deploying AI go way beyond the obvious job savings. At MakoLab, we divide them into three primary areas of real value generation.

A. Growth in revenues and sales 

AI is your best salesperson. It works 24/7, without emotions and without breaks and it’s capable of analysing thousands of clients’ and customers’ needs at once.

·        Advanced recommendation systems in e-commerce: Giants like Amazon and Netflix built their empires on this. This isn’t about ‘Others also bought...” State-of-the-art AI engines analyse the entire history of a client’s or customer’s behaviour in order to propose a product they may genuinely need.
The result? A direct rise in average order value (AOV) as the client or customer quite simply adds more to their cart.

·        Dynamic pricing: Imagine an algorithm that analyses demand, the weather, the competition’s prices and thousands of other variables in order to find the perfect price point. Airlines have been doing it for years. Now AI is making it possible to leverage that strategy in e-commerce and retail.

The result? Margins are maximised without frightening clients and customers away and sell-through rate increases.

·        Intelligent lead scoring: Your sales team has a thousand leads. Which ones are ‘hot’? AI analysis data about opened mails, visits to a price list, e-books downloaded and company names in order to assign a score to every contact. The result? Sales staff only call the most promising leads. The sales cycle is reduced by days or weeks and conversions rise.

B. Reducing costs and turbocharging processes 

In this area, AI operates with forensic precision, eliminating inefficiency, automating dreary tasks and restoring the company’s most precious resource... time.

• Intelligent chatbots / AI assistants in customer service: Forget about the infuriating bots of yesteryear! Cutting-edge assistants rooted in large language models (LLMs) are capable of genuinely understanding and solving clients’ and customers’ problems.

The result? Almost 70% of the routine ‘what’s the status of my parcel?’ and ‘I’ve forgotten my password’ kind of queries are serviced automatically. Consultants are freed up to deal only with the tricky, high-value conversations, while the cost of handling one-off questions drops dramatically.

• Optimising logistics and supply chains: Fuel is a gigantic cost for transport companies. For warehouses, space is at a premium. AI analyses live traffic, weather forecasts, delivery schedules and weight restrictions to propose routes that will save mileage and driver time.

The result? Real savings of 10-15% on fuel and better use of fleets.

• Handling invoices and documents with robotic process automation (RPA) plus AI: Optical character recognition (OCR) and natural language processing (NLP) mean that AI has the capability to understand unstructured text on the scan of an invoice, even if it’s been scanned crookedly or in PDF format. The system itself knows what a VAT number and a net total are.

The result? Around 80% of bookkeeping processes or, in HR, onboarding processes and similar, are automated, minimising the risk of human error.

C. Risk avoidance and intangible value

Not every benefit can be calculated at once in your currency of choice on an Excel spreadsheet. Sometimes, the greatest return on investment comes from the problems that don’t happen.

• Predictive maintenance: This is a change from the philosophy of ‘if it ain’t broke, don’t fix it’ to ‘fix it before it breaks’. IoT sensors on a production line send data to AI, which learns the vibration patterns and temperatures that precede a breakdown.
  The result? Instead of the entire factory suddenly grinding to a halt, the system issues an alert, such as, ‘Replace bearing No. 7 during the servicing scheduled for the weekend’. Impressive sums in savings will accrue.
• Fraud and anomaly detection: This is vital to the finance, insurance and e-commerce sectors. AI analyses every single transaction in real time, searching for patterns that diverge from the norm, working faster and more precisely than people.
  The result? Fraud is blocked before finalisation. This means not only reductions in losses, but also a massive increase in the trust that clients and customers place in your platform.
• Improved prognosis and better strategic decisions: Rather than company strategy rooted in the management board’s ‘intuition’, AI can analyse historical sales data, market trends and social media sentiment in order to forge hyper-precise forecasts of demand.
  The result? The company knows the quantities of goods it needs to order for an upcoming quarter and which markets to enter.
  

Deploying the technology is merely the start. The organisations that achieve the best financial results are those that think strategically, operate agilely and never forget the human factor.

‘Quick wins first’. What is it?

Don’t try to begin by building a ‘Terminator’ that will revolutionise the entire company. Instead of revolution, start with evolution. Pick one process that’s a pain point, but relatively simple, like automated answers to your clients’ or customers’ ten most frequently asked questions. Measure the results after three to four weeks, prove the value and show the management board the concrete figures. Then whet the board’s appetite for a larger project and extract a budget from them. Success breeds success.

It can’t be said often enough; AI mirrors your organisation. If your data is a mess, then all AI will give you is automated mess. Investing in data governance, quality and consistency is absolutely fundamental to ROI.
Poor input data = poor output decisions.  

Classic ROI is ‘here and now’. AI, on the other hand, is a long-term investment. Which is why, at MakoLab, we also look at intangible indicators. How has the deployment affected customer satisfaction vis à vis the net promoter score (NPS) and/or customer effort score (CES)? How has it impacted staff satisfaction and retention, as per the employee net promoter score (eNPS)? A happy staff member and a loyal client or customer make for pure profit in the future.

This is critical. Even the very best of algorithms will fail if your team boycotts it or is afraid of using it. Involve your staff in the process from day one. Show them that AI isn’t the enemy that’s going to replace them, but an assistant, a co-pilot that will relieve them of the worst parts of their jobs, enabling them to focus on creativity and strategy. Make them the ambassadors of change, rather than its opponents.

Investment in artificial intelligence is no longer optional. It’s essential to maintaining the pace in today’s digital economy. Calculating ROI precisely not only makes it possible for you to justify the expenditure to your management board. First and foremost, it serves as a compass, showing which initiatives are working, where further investment will be worthwhile and which projects need shutting down. It enables you to transform technology into a sustainable engine of growth and a genuine competitive advantage.

AI doesn’t replace people. However, when organisations learn to calculate and maximise the ROI it provides, it allows them to take the place of those that don’t.

This would be a perfect moment to dive into our post on the development of chatbots and AI agents.
Visit Digital Solutions Project House | User-Friendly Solutions | MakoLab