AI that saves your staff time without replacing them

Recently we’ve been working on a project that we feel is a perfect example of how AI can be used in a business in a way that’s truly useful, without compromising the human support that customers come back for.

We have all become used to seeing companies introduce AI chatbots and assistants as a way to reduce their staff’s workloads. In some instances these can be really beneficial, especially for larger companies that would ordinarily have longer support response times, with the chatbot acting as triage and helping customers find answers to their questions more quickly.

But for many companies there is a more effective way to integrate AI – one that reduces workload while maintaining the personal, human aspect of the business that their customers value. This project does exactly that.

The setup

Our client runs a comparison service. A customer approaches them with a requirement, and they put it out to a panel of suitable businesses, who each come back with a quote or proposal. Our client then provides their customer with a concise report setting out the options side by side, making the decision-making process much simpler.

The bottleneck

Every business on the panel quotes in its own format, using its own pricing structure, its own terminology, and its own assumptions about what is included as standard and what is an optional extra. The quotes often arrive by different means, too.

Our client needs the same set of information from every one of them, because that is what makes a clear comparison – and a decision – possible.

To get it, a member of their team has to comb through each quote by hand, pull out the relevant details and type them into the comparison report before it can be sent to the customer. It is a highly time-consuming process.

What we didn’t do

There are several ways to approach this problem. Here are two we considered but ruled out.

Writing software to read each business’s format

This is absolutely possible, but it would mean new code every time a new business joined the panel, and breakages whenever an existing one altered its template. The maintenance cost would grow every time our client expanded their panel – the opposite of what they need, given that a broader panel is the point of the service.

Handing the documents to AI and asking for the report

Again possible, and likely fast to build. But for this client it would be a mistake, and would undermine their business model. These reports are customer-facing and commercially significant, so the same inputs must produce the same output every single time. Their customers come to them for their expertise, which means they have to be able to explain precisely why each figure appears where it does. And the rules of the report – how options are presented, what counts as comparable, where caveats apply – are decisions the business needs to make deliberately, so they have to stay controllable and testable.

What we did

We used AI to extract the relevant data, and a piece of custom software, written by us, to build the report.

AI sifts through each quote, plucks out the information the report needs, standardizes it and funnels it straight into a deterministic program – traditional code that produces the same result every time – which then generates the report. This removes the need for staff to read every quote and type the data in by hand, freeing them up to focus on the skills they were employed for.

Data collection is taken care of by AI. Reporting is taken care of by code carefully written and tested by our experienced developers. And our client’s customers are taken care of by their expert staff.

How we ensured the quality of our AI integration

This is a question we’d advise any business owner to ask their developer. As we always say, AI should be thought of as a tool rather than an independently thinking brain: the quality of the output depends on the quality of the prompt. So here’s how we made sure it was reliable.

• A strict, predefined structure. We don’t ask the AI any open questions. It is asked to fill in specific fields, in specific formats, with specific units.
• Permission to leave a field blank. If a quote doesn’t contain a required piece of information, that is recorded in the report. The AI never fills a gap with an assumption.
• A robust testing process. Our AI integrations are subject to the same rigorous testing as our own code – automated tests that run continuously and fail loudly the moment something changes, as well as in-depth human testing by our developers.
• Human checks. Our client’s expert team review every report before it goes out, so any anomalies are picked up.
• Human overrides. No two situations are ever exactly the same, so their staff can add fields to a report manually where a case calls for it.

Effective adoption of AI doesn’t mean removing humans. It means cutting the time they have to spend on repetitive, laborious tasks.

What it has changed

Capacity is no longer locked to headcount, turnaround times are shorter, and experienced staff spend their time on customers and judgement calls rather than transcription. Our client can also gather quotes from more businesses, which improves the customer’s experience without increasing anyone’s workload.

Where else this fits

We actively look for this shape of problem when quoting for new projects, because it turns up everywhere. AI could be used this way in your business if:

• information arrives from external sources in inconsistent formats
• it has to be standardized before it can be used
• it is then processed in a reliable, rule-based way

Supplier quotes and invoices. CVs and applications. Insurance and claims paperwork. Survey, inspection and lab results. Tender responses. Onboarding forms. Expenses. It’s a pattern that turns up in almost every industry.

The unglamorous option

For most of the businesses we talk to, the “boring” behind-the-scenes uses like this one are the most practical way to make use of AI at the moment. It may not be something flashy you can showcase to your customers, but it genuinely improves the experience for you and for them – and that’s a win in our book.

If you are curious about how AI could help save time in your business, get in touch.