EthosPowerAIWhat has changed
Your staff used to know more than your customers: the standards, the sizing methods, the failure modes and which product suits which job. That knowledge earned your margin. Now a customer's own AI prepares the load list, sizes the cables and the battery, names a make and model and lists the suppliers, including your competitors. All of this happens before anyone phones you. When they do call, you compete on price.
Adding AI to the processes you already run does not fix this. It does the same work faster, and once a process is faster, people are less willing to change it. McKinsey's 2025 survey found that 88 per cent of respondents used AI somewhere in the business, but only 39 per cent could point to any profit from it.
Four steps, in this order
- Write the five-year plan. Decide what your customers will still pay you for in five years, when their own AI does the sizing. Move from supplying a product to advising on its whole life. A bank that knows a customer's supply chain can help that customer run it better. A supplier can do the same with the equipment it sells.
- Find the problem that lasts. Choose a recurring cost in your customers' core work, one they will pay to have solved every year.
- Use the data customers trust you with. Test results, SCADA logs and maintenance records sit on private networks, where public AI models cannot see them. A customer shares them only with a supplier it trusts, and your reputation earns that trust.
- Prototype with the customer, then run it. Build a working version in weeks, then test and change it with the people who will use it. After that it runs as a secure service that records what it does, with a person approving each decision. Tell your own engineers first what they gain from it. Nothing I build replaces staff.
A worked example: standby batteries
A network utility runs hundreds of substations. Each has a standby DC system. The charger tests its own battery on a schedule and reports the result to SCADA over DNP3 or Modbus TCP, so decades of results already sit on the utility's own network. Each result bears on the same question: when will this battery fail, and when should its replacement be ordered? If the battery fails first, the switchgear loses its supply. If the order is late, a maintenance outage waits for a delivery.
The supplier who wins this work builds a platform that reads those results and forecasts which strings to replace and when. It checks stock, lead times and prices, schedules the site work and recommends when to raise the purchase order. A person approves each step. The utility gets fewer surprises, and the supplier becomes the firm it plans with, not the firm it calls when something fails. This is hard to do, and most of the difficulty is in the data.
Why it is affordable now
One experienced developer can now build and deploy a platform that once took a funded team years. ebatt.ai, my battery sizing and quotation platform, is an example. It puts IEEE 485 and EN 50272-2 calculations, and the standards checks I have used since 1992, in the customer's hands. Its engineering core is moving to the AGPL-3.0 open-source licence, so you will be able to start from it. Software like this can be copied, so it is no longer the barrier. What a competitor cannot copy is your private data and the trust behind it.
Where this comes from
I have worked this way before, on a small scale. In the late 1990s I gave a small firm's distributors remote access to its ERP system so that they could serve themselves. Later I wrote courses so that engineers in other countries could size standby systems without waiting for me. Since 2024 I have built ebatt.ai. The details are on the about page.
How I work
I do the strategy, the design, the build and the training myself, so nothing is lost between the engineer who understands your business and the one who writes the software. I work from Nelson for clients across New Zealand and Australia. The prices are on the services page. Everything I build for you is open source and can run entirely on your own servers, so your compliance team can read the code and you keep it when the work ends. The 35 platforms I use are on their own page.
The limits of this argument
Afanasyev's account comes from practice. It rests on two recorded interviews, not a study, and his employer sells cloud services. My own evidence is the three cases described above. You can test the idea cheaply: ask your next three customers what their AI told them before they called you.