← back to blog

The trust wall between businesses and AI automation

#ai #automation #business

I started talking with businesses to see which problems neonloops solves in the real world. I’m continuing these conversations with different companies in accounting, finance and education.

I wanted to share what I’ve seen so far.

How the meetings go

In short, I knock on businesses’ doors and tell them I can set up AI-powered automations for them, and AI workflows that run their current processes fully automatically. I tell them I’m not selling anything, and that I’m only there to listen to their needs and answer their questions if they have any.

When the topic is AI and automation, I’d say most of them are interested. Out of curiosity about what the new age brings, and to see what’s possible, every place I’ve gone to so far has agreed to meet me. In the meetings, I focus less on what I could solve with automation and more on listening, so I can understand what the person in charge at the business might need.

I always start with a short introduction of myself. Then I tell the person I’m curious about what they might need and that I want to listen to them. I’ve noticed that most of them don’t have enough experience with or knowledge of what AI can do. The first questions usually come from not understanding how an AI could automate the workflows they have. They explain to me how complex their processes are, and tell me they employ people for this, and that even training those people takes a lot of time.

Then, with a few questions, I try to understand why an AI couldn’t do it. In the explanations that follow, they say they don’t know whether the rules in their workflows and the tools they use are something an AI could work with. Let me give an example.

Ahmet’s two questions

I had the chance to meet one of the partners at an accounting firm with 15 employees. Let’s call him Ahmet. As I described above, I introduced myself, then asked which jobs took up most of their time and how often those jobs repeat.

Ahmet said most of his employees’ time goes into the flow where invoices and receipts get entered into the accounting program, because the information on the physical invoices first has to be sorted by account codes from a chart of accounts and saved in Excel. After that, the employees take these separate Excel files, one for each client business, and enter the information into the accounting program they use. Finally, he said he didn’t understand how an AI, after reading an invoice, could tell whether the items on it are an expense or not, depending on the type of business. And even if it could, he didn’t understand how it would write those expenses into the column with the right account code in an Excel file.

I do a lot of complex work with AI myself, so I knew both questions had simple answers. I asked him whether ChatGPT could do this if he asked it. Ahmet got it, and realized this was possible even with today’s AI chatbots.

After that conversation, before I left, I told Ahmet that any digital flow they have can be automated by AI. I even told him AI can already do this work better than people.

A demo with their own invoices

As the example above shows, processes that businesses think aren’t possible yet are already possible with AI. To prove it, I prepared an accounting demo, and I showed it in a meeting with another accounting firm.

The demo covers the process that takes up most of an accounting firm’s time: invoices and receipts. You photograph them all in one go with a phone and press a single button. Automation then runs the whole process for that business and turns it into an output that fits whatever accounting program they want. I showed it working right in front of them.

The person I met was really surprised. They expected this to take very advanced software with a whole team working on it, but in this sample interface, running a workflow I designed on neonloops, every step came out right. I even asked for a few of their own invoices for the demo, and when I ran it on those different invoices, they couldn’t find anything to say against it.

But afterwards, they said they didn’t know whether this solution meets accounting standards, or who would be responsible for mistakes it causes. They also said the big companies that sell accounting software have similar processes, but none of them could build a workflow that works this well.

The trust wall

From what I understood in that meeting, even when an AI solution works very well in practice, businesses aren’t very willing to automate their existing workflows, because they’re looking for something to base their trust on. So I see a trust wall. But they don’t realize that in a future where all these solutions are common, they’ll pay much more to get access to them, and they’ll have to replace their existing workflows with systems forced on them.

So I noticed that businesses that are very interested in AI and automation, in the end, look for a regulation or a rule they can hold on to, even when it’s possible to apply the solution and benefit from it. I’ll be looking into ways to give that assurance. Still, in the move to the AI age, before we reach a time when every business’s processes are automated, it isn’t hard to predict that the businesses that make the leap on this will succeed. As with every new thing, the ones that take the risks and accept the mistakes will maximize the benefits that come from those risks.

And I’ll be sharing what these processes bring in this age of transformation.