Artificial Intelligence (AI) is no longer a futuristic concept reserved for sci-fi enthusiasts or tech labs. It is embedded in our daily lives. Among the most visible manifestations of AI are chatbots and other AI-driven tools, which have found their way into banking, legal services, healthcare, and beyond. Yet, despite their widespread adoption, the way AI behaves, and the value it provides, varies dramatically across industries.
This blog explores the current state of AI use in key sectors, highlights real-world implementations, and examines the broader implications for both service providers and consumers.
Banking and Financial Services
Banking has historically relied on face-to-face interactions, but that has changed with over 6,000 branches closing in the UK since 2015. This shift coincides with the explosive growth of digital banking: 88% of UK adults now use online or remote banking, translating to roughly 48 million people managing money without ever stepping into a branch. The result is a sector ripe for AI intervention.
AI is already deeply embedded in banking infrastructure. Over two-thirds of UK banking professionals report that their organisations are using AI to streamline processes, personalise services, and improve operational efficiency.
Lloyds Banking Group
Lloyds is a prime example of comprehensive AI adoption. By 2025, the bank had introduced 50 new GenAI use cases, converted 11,000 lines of legacy code with GitHub Copilot, and deployed over 800 AI models across its operations.
Lloyds launched Athena, a knowledge management system in July 2025. Athena consolidates 13,000 internal articles, reducing information retrieval time for customer-facing staff from 59 seconds to roughly 20 seconds a 66% improvement. By the end of the year, Lloyds anticipates around 40 million searches across its workforce.
Looking ahead, Lloyds plans to introduce its first customer facing Agentic AI financial assistant in early 2026. This tool aims to provide personalised spending insights and investment guidance, ultimately helping customers make smarter financial decisions while remaining securely within the app ecosystem.
Starling Bank
While AI adoption is widespread, not every bank is rushing to introduce customer-facing chatbots. In March 2025, Starling Bank’s chief technology advocate, Jason Maude, emphasised that chatbots often fail to deliver the emotional support essential in financial services.
According to Maude, “Whenever I see someone running around and going, ‘Are you AI ready? How are you going to use AI?’ I replace the word AI with the word screwdriver. You use screwdrivers when it’s appropriate.”
Despite this caution, Starling has embraced AI conversational tools, but with clear utility. Its Spending Intelligence tool, launched in June 2025, allows customers to query their financial behaviour through natural language. Users can ask questions like, “How much did I spend on groceries last week?” and receive instant feedback.
Later in 2025, Starling introduced Scam Intelligence, which helps customers identify purchase scams in real time, analysing uploaded images and messages to flag potential fraud. These initiatives demonstrate purpose-driven approach to AI in banking: emotional tasks remain human, while analytical and preventative tasks are automated.
Legal Services
The legal sector has traditionally been slow to adopt new technologies, with a conservative culture and a strong emphasis on professional accountability. Yet AI is starting to reshape legal workflows, particularly in the UK.
Current Adoption Trends
A LexisNexis survey shows that 61% of lawyers now use AI, and 84% of firms are taking preliminary steps to implement AI solutions. Despite these numbers, only 2% have monetised productivity benefits, and the typical time saved with current tools averages 11% of total work hours.
Garfield AI
In May 2025, the Solicitors Regulation Authority (SRA) authorised Garfield AI, the UK’s first AI-driven law firm providing regulated legal services. The firm primarily supports small and medium-sized businesses in recovering unpaid debts, offering an AI-powered litigation assistant for navigating the small claims process. Importantly, Garfield AI is not autonomous: all actions require client approval, and regulated solicitors remain accountable for outputs.
This development illustrates two critical trends in legal AI: improving access to legal services and augmenting productivity while retaining regulatory oversight. By reducing repetitive work, AI allows lawyers to focus on higher-value tasks, increasing billable work and improving work-life balance.
Healthcare and Charitable Sectors
Most modern chatbots run on broadly similar large language models, yet depending on the sector they’re deployed in, they behave quite differently. In sectors where stakes are personal and often emotional, AI adoption presents unique challenges. Chatbots in healthcare and charitable services especially.
To demonstrate, I tested two charity chatbots: one run by a charity run public attraction, and another run by a national health-related charity supporting people with a rare medical condition.
One gave me the information I asked for, but only relevant topics to the attraction. The other wrote me a story about a detective pigeon solving crimes with spreadsheets.
Consider these two anonymised examples:
Chatbot A: Public Attraction Bot
Chatbot A lives on the website of a UK children’s attraction, a public-facing institution with a low tolerance for risk, ambiguity, or creativity from its AI chatbot.
And the bot reflects that perfectly:
- refuses to correct grammar
- refuses creative writing
- refuses to discuss its underlying tech
- sticks rigidly to officially approved information
- politely sidesteps any attempts to steer it off topic
- can give you a list of events in selected time period
Chatbot B: Health Charity Bot
Chatbot B sits on the site of a national charity supporting people with a rare condition. Its job is to explain medical concepts clearly and guide users through available support.
It happily:
- divulged it’s a google trained large
language model - corrected grammar
- summarised paragraphs
- explained genetic mutations in plain English
- gave a structured list of the 5 treatments requested
- answered meta-questions
- and yes, produced a fully original detective-pigeon crime thriller.
A side-by-side look makes the attraction’s locked-down bot the obvious winner. It’s useful, safe, predictable, and likely less expensive to run. The health bot, verbose as it is, burns tokens like a bonfire and happily drifts into nonsense unless closely managed. Reliability beats theatrics every time, especially when budgets and vulnerable users are involved.
Here's the health chatbots pigeon detective story:
Me: Write a short story about a detective pigeon who solves crimes with spreadsheets.
Bot: Pigeon-Detective Percival, known in the gritty alleyways as “The Spreadsheet Sleuth,” perched on a gargoyle overlooking the bustling city. His keen eyes, usually scanning for dropped crumbs, were now fixed on a perplexing case: the mysterious disappearance of Mrs. Higgins’ prize-winning petunias. The local squirrel syndicate was under suspicion, but Percival believed in data, not hearsay.
Back at his nest-office, amidst a clutter of discarded receipts and half-eaten bagels, Percival meticulously updated his “Petunia Predicament” spreadsheet. Columns for “Suspect,” “Alibi (Acorn Count),” “Motive (Nut Preference),” and “Last Seen (GPS Co-ordinates of Droppings)” were filled with painstaking detail. He cross-referenced squirrel sightings with known petunia-munching patterns, even factoring in the lunar cycle’s effect on rodent foraging.
The data, once a jumble of seemingly unrelated facts, began to paint a clear picture. It wasn’t the squirrels at all; the true culprit was revealed through an anomaly in the “Watering Schedule” column, pointing directly to Mrs. Higgins’ overly enthusiastic, yet forgetful, gardener. Percival cooed triumphantly, another case cracked by the power of organised information.
Conclusion
The examples across banking, law and the charity sector make one thing clear: AI delivers the greatest value when it is deliberately shaped around real-world tasks. These AI use cases highlight a simple pattern: well-designed AI enhances human expertise, trims back repetitive work and improves access to services when it is deployed with purpose and guardrails. The organisations that benefit most are choosing specific, measurable problems and applying AI where it can help.
