AI: My Top 4 Use Cases in Client-Specific Digital Solutions

In the early days of AI, called machine learning, I got involved early and built some useful applications. One was a fraud detection algorithm for advertisers, among other things in finance and media. But back then, you had to put a lot of work into preparing the data, a process known as supervised learning. You had to label the information based on human experience. Training those models was time-consuming, difficult, and expensive. Few organisations had the right data to start with, let alone the ability and resources to label it or tag it to teach the machine.

Then chatbots came along, based on that technology but specialised. They were even more disappointing. Anyone who tried to use that technology soon realised its limitations in human interaction. It was more about vendor hype and consulting hype because you could make a lot of money putting those things in and configuring them. They tended to be more expensive to configure than it would have been to just build a complex bit of software to do the same thing.

The biggest Achilles' heel of those chatbots was their fragility. Over a few months, they would often stop being effective. Humans would adapt and use different phrases, but the machine was not so flexible. In practical scenarios, they tended to degrade over time. So you'd constantly be working and reworking them, configuring and reconfiguring them, which was very expensive. Most of the people I know in big corporations tended to move away from chatbots once they stopped being initially useful.

There have been maybe 10 fads since then. But GenAI is different, and I'm genuinely excited about the wide variety of situations this technology can be applied to with relative ease. Let's look at some examples that might help you spot opportunities to apply the technology in your own organisation.

The first area I would characterise as getting answers instead of search results. We've all tried searching for something on Google, a specific website, OneDrive, or SharePoint, only to get a thousand results. You then have to sift through these results, which are not answers but links you must read and synthesise to find the answer, which is inefficient.

A recent example: I searched an education charity's website, and every query returned about 14,000 results. The top few search results were always the same, making them useless in answering the question. You could click through the whole first page of results and still not find what you were looking for. This is really common. Those search results couldn't answer basic questions like how to apply, volunteer, invest, contribute, or donate.

What we did was put all that content into a specialised AI agent. Instead of the usual search results, we immediately got an actual answer to the question. We manually loaded the website's content (all publicly available information), asked the same questions, and got real answers. The answers could then link to the documents, successfully navigating the noise of search results. So you get an answer plus links to read if you want.

I've seen this applied internally with OneDrive and SharePoint. We've all experienced searching for something, not finding it, and not finding the answer. A lot of knowledge exists on OneDrive and SharePoint, and using it more effectively would be amazing. The same solution works here: ingest the content into a AI agent, ask a question, and it will answer based on the drive's information rather than public data like ChatGPT does. In many situations, you need to constrain it to a specific knowledge base to get the answer you need.

I've even seen this work with videos. Automatically transcribed, you can ask a question based on interviews across, say, 10 videos, and the agent can mark the specific time in each video for you to listen to the original transcript. It isolates the information you're looking for and gives you an answer rather than a search result. This is the easiest and most prevalent pattern of problem I see going around because everyone has a search function somewhere.

The second area is summarising large volumes of information. As a human, you need to read and digest a vast amount of knowledge to create your summary, which is tough. Most charities have valuable information about customers and internal processes, but summarising it is challenging due to the effort required.

AI excels at searching for information and providing answers instead of search results. It can also summarise the information it finds. This can greatly improve the quality of life for staff and the service offered to customers. For example, trawling through a client's unstructured case notes from different formats to get a summary of the last two years' activities is very difficult. But with a AI agent, it's very easy. It will summarise the information and can even prepackage it in another field in your client management system.

Another useful application is summarising the humans involved with a client in different capacities. This involves aggregating information in various formats into a central area, similar to data warehousing but with unstructured information.

I've seen this done with single documents, like a tender document, to summarise what the contract covers without reading the entire 80-120 page document. This is powerful, especially for legal corpora, tax accounting rules, and similar documents. Even as a lead-in to reading online materials, this can save a huge amount of time by directing attention to specific areas in the summary.

The third area is GenAI's ability to hold much more human-like conversations. Previous chatbots can't compare to the power of AI in this regard. It’s nearly human, capable of unstructured conversation while still capturing structured information in backend systems. For example, you could talk to a mobile app instead of typing information into a client management system. While driving, you can chat away, and AI can record, transcribe, summarise, and structure the information in the system.

This also works with messaging platforms like SMS, allowing for back-and-forth conversations that feel like there's a human at the other end. It can prompt you for missing information, make suggestions, record instructions, summarise conversations, and input data into systems. This combination of conversational capability and backend management will see many advancements in the next year.

The final area is translation from one language to another. Previous attempts were disappointing, but AI's nuanced translation offers new opportunities. Workers can record case notes in their first language and have them translated into English for the client management system. Storing the original audio or casework is also possible, which is very powerful.

For example, one of my recent clients had a worker for whom English was the seventh language. Writing useful case notes in English was challenging, but writing or speaking them in her first language and having them translated would have been easier. AI has changed the game in this space, making things practical for the first time.

These four areas are worth considering when looking for problems and opportunities in your organisation. My experiences with AI often remind me of Google CEO Sundar Pichai's AI-first policy from 2015 or 2016. He required internal teams to attempt a machine learning solution before spending significant money on software development. This policy resulted in breakthrough technologies, especially in cloud services.

I have a similar approach with AI. I start with that as my first port of call, reverting to traditional methods if necessary. I won't ask a developer to build something or even create spreadsheets in many situations. I'll give information to a specialised AI solution and let it work for me. AI is revolutionising our operational lives in client-facing admin roles, jetpacking software engineers and non-software engineers alike. From personal experience, I can say it's incredibly fun. So, give something like ChatGPT a go, explore what it can do with public information, and think about how to apply its strengths in your daily life.

Andrew Walker
Technology consulting for charities
https://www.linkedin.com/in/andrew-walker-the-impatient-futurist/