I thought I was pretty good at AI.
I use it every day across my businesses for writing, research, ideas, analysing information, working through problems and all sorts of things that I would previously either have spent hours doing myself or put on a list to do later, where they would have remained until approximately 2047.
I can give it a spreadsheet and ask what I’m missing, throw a slightly chaotic collection of thoughts at it and ask it to make sense of them, or use it to interrogate an idea until I have worked out whether it is actually a good one.
So, compared with somebody who thinks AI is basically ChatGPT writing an email beginning “I hope this message finds you well”, I felt I was doing rather well.
And then I spoke to Neil, founder of Codefully, and realised that while I might use AI a lot, I’m still largely using it in a very sporadic way.
I have a question, I ask it. I have some data, I give it the data. I need to understand something, I ask it to explain it.
All incredibly useful, but what it doesn’t necessarily do is connect all the moving parts of my businesses together.
I run relatively small businesses, but even at my size there are accounts, invoices, sales, leads, clients, marketing, social media, staff, projects, content, follow-ups and all the little operational things that somehow breed overnight.
There is information about almost all of it, but that information lives in different places and arrives at different times.
One person knows this, another spreadsheet contains that, the accounting software knows something else and somewhere in my inbox is probably a very important email that I definitely intended to reply to three Tuesdays ago.
What I actually want as the person running the business isn’t more data. I have plenty of data.
What I want is the helicopter view.
I want to know what is going well, what isn’t, what has changed, what is costing too much, what we haven’t followed up, what needs my attention and, preferably, what I should do about it.
I don’t particularly want to open seven systems and compare them myself, and I definitely don’t want somebody in my team spending half a day producing a beautiful 14-page report only for me to look at page one and ask, “Yes, but is this good or bad?”
I want the useful bits, regularly, consistently and in a way that allows me to make better decisions.
That was the moment my conversation with Neil became genuinely interesting, because he wasn’t trying to sell me on the idea that AI is clever. We all know AI is clever.
He was making me think about whether I am asking it to do the right things.
We May Be Using AI Completely Backwards
Codefully is an AI implementation agency with offices in London, Dubai and India, building custom AI solutions for businesses.
But Neil’s starting point isn’t to walk into a company and start throwing AI at everything that moves.
In fact, his starting point is much more ordinary.
He looks at how people actually spend their days and asks where the friction is.
What are they doing manually? What are they doing repeatedly? What information are they copying from one place to another? What are managers constantly chasing? What reports take hours to produce? Where is useful data sitting there doing absolutely nothing useful?
“Most companies already have AI tools sitting unused,” he explains. “What we focus on is finding the manual, repetitive parts of a team’s day and building solutions around those specific problems, ones that actually get used and make a measurable difference.”
And this is where I realised that I have probably been approaching AI in the same way as a lot of business owners.
We hear about something new and ask, “What can I use this for?”
Neil effectively turns that question around and asks, “What is driving you mad?”
That could be somebody entering the same customer information into three systems, a manager spending Friday afternoon compiling figures for Monday morning, a sales team forgetting to follow up leads, or the business owner who knows they have all the information somewhere but still has to ask five people to find out what is actually happening.
Instead of starting with AI and looking for somewhere to put it, start with the problem and work backwards.
That sounds obvious once somebody says it, but it is quite a fundamental difference.
If I sat down and listed every repetitive process across my own businesses, every report I would like to receive automatically, every piece of information I regularly have to ask somebody for and every occasion on which I discover something later than I would have liked, I suspect the list would become alarmingly long.
And I am a small business owner.
Now imagine doing the same exercise in a company with 500 people, or 5,000, where hundreds of employees are carrying out repetitive tasks every day, departments are operating on different systems and senior management is trying to make decisions using information being produced in multiple places.
Saving one person 30 minutes a day is nice.
Saving 1,000 people 30 minutes a day is something the finance director is going to become considerably more excited about.
We Don’t Have a Data Problem. We Have a “What Does Any of This Actually Mean?” Problem
The irony is that most businesses already have enormous amounts of information.
Even mine do.
We have sales figures, accounting data, social-media analytics, website information, email performance, customer details, invoices, enquiries and all sorts of reports.
Larger organisations have exponentially more.
But possessing information and being able to use it intelligently are completely different things.
You can have the most beautiful dashboard in the world, but somebody still needs to look at it and understand what matters.
You can have ten reports arriving every Monday, but if nobody joins them together, you might miss the fact that marketing enquiries have risen while conversions have fallen, or that sales look healthy but overdue invoices are quietly stacking up behind them.
You can know what happened last month without understanding why it happened or what you should be watching next month.
For me, this is where AI becomes much more exciting than simply producing things faster.
Imagine the information across a business being pulled together so that instead of constantly going looking for it, the relevant information finds you.
Imagine receiving a useful management summary that doesn’t simply tell you that sales are £X and enquiries are Y, but flags that something unusual has happened, identifies where attention is needed and gives you enough context to ask the right question.
I don’t need 47 graphs.
I need to know that one number has suddenly gone the wrong way and nobody has noticed.
And again, take that idea to scale.
In a large company, the value isn’t simply the hours saved producing reports. It is the potential value of seeing things earlier.
A pattern in customer complaints, a bottleneck in an operational process, duplicated work across departments, a change in buying behaviour, an area where costs are creeping upwards or an opportunity sitting inside data that nobody has had the time to connect.
At that point, AI isn’t simply saving somebody an afternoon of admin.
It is potentially changing the quality and speed of decision-making.
What If Your Best People Actually Got to Do the Jobs You Hired Them For?
One of Neil’s points that resonated most with me was his view that the best use of AI isn’t necessarily replacing people.
It is removing the parts of their jobs that people probably shouldn’t be doing in the first place.
“I believe the best use of AI is taking the repetitive stuff off people’s plates, the admin, the data entry that gets redone across three different systems, the status updates people are chasing down over email, so people can actually spend their time on the parts of the job that need real judgement and are exciting to them.”
I love this way of looking at it because every business has people doing things that are technically part of their job but aren’t really the reason you employed them.
You have somebody who is fantastic with clients spending an hour updating a spreadsheet.
A creative person copying information between systems.
A senior manager chasing people for status updates.
Somebody with years of experience spending Friday afternoon producing a report that could probably be produced automatically before they’ve finished their first coffee.
Neil sees this as a waste not simply of time but of talent.
“So many companies have really great talent who aren’t ultimately doing really great work, because they’re stuck with admin, stuck with the things that take up a lot of time, that are repetitive, that are manual, and they leave the office at the end of the day feeling unmotivated, because they’ve spent it doing the things they shouldn’t be doing.”
That is a much more appealing AI conversation than the slightly terrifying one about which jobs are going to disappear.
Perhaps the first question shouldn’t be, “Which people can AI replace?” but “Which bits of people’s working days can we give them back?”
If AI can take away the repetitive administration and leave people with the work that requires relationships, judgement, creativity, empathy and experience, you potentially end up not only with a more efficient company but a happier one.
And again, scale it.
Ten minutes saved here and twenty minutes there sounds insignificant until you multiply it across thousands of tasks, hundreds of employees and 250 working days a year.
Suddenly all those little inefficiencies that everyone simply accepted as “part of the job” become a fairly enormous amount of money.
And Can We Please Talk About Meetings?
Codefully’s own product, askInput, came from one of the most relatable examples of this thinking.
We have all attended meetings that could have been emails, but there is an even more irritating category: meetings that exist purely because somebody needs information from several people.
First there is the email asking for everyone’s availability, then the inevitable 19-message exchange because Dave can’t do Wednesday and Priya is away on Thursday, eventually a slot is found and six people spend an hour on a call when the actual purpose was to ask each of them a handful of questions.
Codefully looked at that process and essentially asked: why are we having the call?
askInput collects everyone’s input, allows participants to see what other people have contributed and then uses AI to collate and summarise the responses into meeting notes or even a complete document.
The clever bit is obviously the technology, but from my point of view the really clever bit is that six people just got an hour of their lives back.
Multiply that across a large organisation and suddenly it isn’t a funny little observation about too many meetings.
If ten people attend a one-hour meeting that didn’t need to happen, the company hasn’t lost an hour. It has lost ten.
If versions of that meeting are happening hundreds of times across an organisation every month, the numbers become eye-watering.
Of Course, There Is One Slight Problem: You Have to Trust It
The deeper AI goes into the running of a business, the more serious the questions become.
I’m perfectly happy asking AI to suggest a better headline.
Giving it access to operational data, customer information and business processes is obviously another matter entirely.
Neil believes this is actually one of the biggest barriers to meaningful AI adoption.
“The biggest challenge, honestly, isn’t technology, it’s trust,” he says.
Businesses want to know whether a system will work, whether it will fit into existing processes, where their information is going, how data protection and data residency are being handled and whether what they build will remain compliant as regulation evolves.
“It’s less about whether we can build the AI and more about whether we can build it in a way that people actually trust and that fits within the rules.”
That distinction matters because the deeper AI becomes embedded in decision-making, the less room there is for the sort of cheerful confidence with which AI occasionally tells you something completely wrong.
If an AI system is going to interact with company knowledge, analyse important information or eventually allow agents to carry out processes, businesses need governance and traceability alongside the clever technology.
Neil believes that will become increasingly important as AI progresses from responding to individual prompts to carrying out more complex sequences of work.
The Bit That Really Made My Ears Prick Up: An AI Brain for the Business
Neil believes the direction of travel is towards something much more connected than the AI most of us are currently using.
Instead of a collection of separate tools bolted onto different departments, he sees businesses developing something closer to a central AI brain, a layer that understands the broader context of what is happening across the organisation.
Underneath that, specialist AI agents could perform specific roles, one dealing with customer queries, another carrying out compliance checks, another scheduling, another analysing information, while an orchestration layer determines which agent should do what and in which order.
Codefully is already experimenting with this internally, and it is now developing an AI brain that companies can integrate their business onto.
Neil describes the concept as giving the organisation its own library and its own AI librarian, which is probably the simplest way of explaining why it could be so powerful.
Most companies already possess huge amounts of knowledge.
The problem is that it is spread across documents, systems, emails, reports, databases, presentations and, rather inconveniently, people’s heads.
The information exists, but finding the right piece at the right time can be another job in itself.
The idea of an AI librarian is that both the people working inside the company and the AI agents carrying out work could access the information and context they need from one connected knowledge base.
For me, running small businesses, that immediately makes me think about all the times somebody asks, “Where’s that document?” or “What did we agree with them?” or “What did we do last time?” and the answer involves searching an inbox, WhatsApp, a folder and possibly the memory of somebody who happens to be on holiday.
At enterprise scale, that same problem is magnified thousands of times.
Institutional knowledge is incredibly valuable, but only if the institution can actually find it.
This Is Where I Realised I Had Been Thinking Too Small
I went into my conversation with Neil thinking about how AI could make my businesses more efficient.
I came out thinking much more about what happens when you apply the same principles at scale.
For me, having the right information automatically pulled together could save hours, improve follow-up, give me better oversight and stop things slipping through the cracks.
That would genuinely make a difference to my working life.
But take the same idea into a business with hundreds or thousands of employees and the mathematics changes dramatically.
A process shortened by ten minutes, repeated 50,000 times a year, is no longer a minor convenience.
A report automatically produced across 20 departments isn’t simply an admin saving.
A pattern identified earlier in millions of pieces of customer data could influence a major commercial decision.
A piece of company knowledge instantly available to thousands of employees rather than rediscovered repeatedly has a value that is much harder to quantify but potentially enormous.
And perhaps most importantly, the value isn’t only in doing the same things faster.
It is in seeing things that humans simply struggle to see when information is fragmented across a large organisation.
Humans are very good at judgement, relationships, instinct and context, but none of us can simultaneously read every customer interaction, analyse every operational report, monitor every sales movement and remember every document a company has ever produced.
Properly implemented AI potentially can connect far more of those dots and then put the useful insight in front of a human who can decide what to do with it.
That is a very different proposition from asking ChatGPT to make an email sound friendlier.
So Maybe I Don’t Need More AI After All
One of Neil’s biggest lessons from building Codefully has actually been about resisting the temptation to become too enamoured with the technology itself.
You can have a brilliant idea and enormous passion for what you are building, he says, but ultimately the customer decides whether you have solved something worth solving.
“Put the client first, put the customer first, put what it is they want to solve first, and you’ll end up with an amazing outcome and an amazing experience for the customer.”
I think that is probably the best advice for any business wondering what to do with AI.
Don’t start by buying another tool because somebody on LinkedIn has announced that it will “10x your productivity”.
Start by looking around your own company.
Look at what people are repeatedly doing, what you’re constantly chasing, what takes far too long, what gets forgotten, what information you wish you had sooner and what data you’re collecting without ever really turning into insight.
Look at the meetings that don’t need to happen, the reports that take hours to create and five minutes to read, the people doing work well below the level of talent you’re paying them for and the decisions you could make better if somebody simply joined the dots.
That is what my conversation with Neil changed for me.
I still use AI every day and I will almost certainly continue asking it all sorts of random things as and when they occur to me.
But I now realise that this is probably the least interesting part of what AI could do for my businesses.
The much bigger opportunity is to stop treating AI as something I occasionally go and talk to and start thinking about where it could sit inside the business itself, connecting information, automating the repetitive bits, spotting what matters and giving me the helicopter view I am forever trying to create manually.
And if that can make a meaningful difference to a business of my size, imagine what happens when you multiply it by 100, 1,000 or 10,000 employees.
Suddenly, saving ten minutes isn’t really about ten minutes at all.
About Neil and Codefully
Neil is the founder of Codefully, an AI implementation agency with offices in London, Dubai and India, helping businesses move beyond using AI for individual tasks and towards integrating it into their everyday operations.
From automating repetitive processes and connecting business information to developing custom AI solutions, Codefully focuses on solving practical business challenges, improving efficiency and helping teams spend more time on work that matters.
To discover how Codefully could help your business, visit www.codefully.ai.
