I swapped a PC for a laptop! Why?
So, I can sit in front of the TV at night, with my wife, and sill work on my business.
Doing those tasks I’ve haven’t been able to get done. The ones that have been on my to-do list for an entire week, even longer!
The owner of an accountancy firm I had a discovery call with had a similar story.
On a Friday night. She grabs her laptop. And sits in front of the TV to get one job done.
Invoice reconciliation.
Even if it takes her till 2AM. She can catch up on sleep over the weekend. Otherwise, month end becomes a nightmare.
She reached out to me to see if I could help to automate this job.
We developed an AI powered automation. That looks at the line item of each invoice. And puts it into one of three buckets, ‘Verified’, ‘To Update’ and ‘Unknown’.
The Hidden Time Thieves in Accounting
A 2025 industry report notes that. 56% of finance professionals spend more than 10 hours per week on manual invoice processing.
That’s 40 hours a month. A full working week.
Over a year that’s a HUGE 12 weeks. An entire quarter of the year lost to one single task.
Do you and your team spend an entire week each month getting it done and out of the way. Or do you set aside a single day each week?
But you and your staff are only meant to be working 8 hours a day. So that’s 2 hours of overtime.
No wonder my client did it herself on a Friday night.
Throw in the fact that the UK Government and HMRC are always moving the goalposts. You might need to submit returns for certain clients every 3 months.
Each accountancy firm is different. But what impact does this task have on other areas or your business?
- Client onboarding isn’t the priority.
- There’s less time to speak to clients.
- Business strategy goes out the window.
More time on this task. Means more costs, in both time and money in other parts of the business.
The True Cost of Traditional Reconciliation
The numbers shocked me.
So, I dropped my own accountant a call.
He is a “one-man-band” nowadays. But he was telling me he and his old business partner used to run a large accountancy practice.
Explaining that each member of staff equates to a set number of billable hours each year. In his experience, this was 1500 billable hours a year.
Most of his time was tracking where this billable time went. On what job and moving things around.
Not growing his business.
Wild!
So, let’s take the average of 10 hours a week. Multiply that by 52 weeks. That’s 520 hours a year.
You’ve wiped out over a third of the billable hours with a single task. Leaving only 980 left for everything else.
I don’t have any stats for this, but can I ask you a question?
What impact does the above scenario have on staff retention? Does your team suffer burnout during certain times of the year?
What about client satisfaction? Do you notice that you get more complaints during specific periods? Or are you noticing it’s dropping across the board? Can you deliver what you promised?
Finally, what about you?
Is running your business impacting your personal life, family time? Your wellbeing and the wellbeing of your family?
Why AI Invoice Reconciliation Now?
Are you asking this question “There must be a reason why it’s still a manual process”?
My view is that we haven’t had any good tools that can understand context.
But that has all changed because of AI.
I started coding automation in 2007. These have always been a ruled-based system. With a very limited “intelligence”
Because of these rules. Systems are:
- Hard to maintain, constant updates to the rules.
- Not very flexible, would look for certain phrases or words.
- Not accessible to all businesses, development was and is expensive.
Now, because of LLM’s, large language models. This has changed.
These models can process and understand text. Understand context. Are trained on real-world data so know what an invoice is. What a line item is.
They can understand if the description on a line item relates to code, an account type.
Now, we can automate the process.
The Three-Bucket AI Invoice Reconciliation System
OK. Let’s get to the fun part. The solution.
Remember the lady I mentioned? The accountancy firm owner. Who worked till 2AM every Friday night.
After chatting with her. And setting her up with a dedicated automation server. I then linked it to her accountancy practice software. And configured the AI Agents as I went along.
I came up with this methodology.
Could we use an AI Agent inside an automation, a workflow. To put an invoice and invoice line item into one of three buckets:
- Bucket 1: “Verified” – transactions processed with complete confidence.
- Bucket 2: “To Update” – transactions requiring minimal human review.
- Bucket 3: “Unknown” – new patterns requiring human decision.
The great thing about these 3 buckets. Is bucket 1, verified, is always the “fullest”. But it requires ZERO human attention.
It’s the other buckets, the ones with the least “water” in that needs the time allocating to.
For bucket 2, the “To Update” bucket. It needs someone to review. To “spot check”. To make sure that the AI Agents suggestion to change the account code is valid.
The third and final bucket, “Unknown”. This is for line items where the AI Agent was unsure of what to do. Don’t force it to make a decision. Pass this on to a human.
Real-World Example
How did I come up with this methodology?
Again, chatting with my client. She mentioned for one of her customers. 99% of their invoice allocations were correct.
For one supplier. That sent an invoice in monthly. They had multiple line items that needed allocating to different account codes.
That client also had one or two custom account codes. That specific items needed allocating too.
The limitation of her current system. Meant the entire invoice got allocated one account code.
Every month, she or a member of her team would have to review each line item on the invoice. And work out what line items to update.
Now, with our AI-powered solution. It produces a report suggesting what to update the code to for specific line items.
What’s crazy about this is the report. Is it includes the reasoning on why the AI Agent is suggesting the changes.
All that’s needed is to spot check the reasoning and new code and update the system.
Insane!
Time Savings By The Numbers
So, what is the result of this automation?
After testing this with a few clients of the accountancy firm
85-95% of line items fell into the “Verified” bucket.
On average 5%-10% of line items dripped into bucket 2, “To Update”.
Only a few occasions the AI Agent did not know what to do with a line item. I fixed this by improving the prompt(s) the AI Agents used.
So, from these results. Let’s take a conservate number that 80% of all line items fall into the “Verified” bucket.
Now, going back to my research. That on average, 10 hours of a working week is spent on manual invoice processing.
Let’s reduce this number by 2 hours.
Why 2 hours? Well, the accountant I’ve been working with has systems in place. Which automates the uploading of invoices into their accountancy software. So, some of the manual work get done on autopilot.
The new numbers is 8 hours a week, a full workday.
That’s 480 minutes.
Now, as our AI-powered automation can verify a conservate 80% of line items. They don’t need any human intervention. Or there’s no need to spend ANY time on these line items.
Only 20% of our time needs to go into review buckets 2 and 3.
Let’s run the numbers.
480 divided by 100, then multiplied by 20 equals 96.
That’s 96 minutes a week.
An 8–10-hour job now only takes to 1 hour 30 minutes to complete.
WAHOO!
Impact On Time & Budget Allocation
Let’s take this time saving and apply it to my accountant’s old business.
Remember, he allocated 1500 hours a year to each staff member. These hours were billable to his clients.
We mentioned that manual invoice reconciliation took 520 hours a year (10 hours a week).
Let’s reduce that to 8 hours a week (because of automated invoice uploading). That’s 416 billable hours gone on a single task.
Our automation has reduced this number down to 96 minutes a week. Let’s say it’s 1.6 hours a week for ease.
1.6 hours a week, multiplied by 52, equals 83.2.
I’m going to round this up to 84 hours. In fact, let’s put in a 6 hour buffer and make it a nice round 90.
With our AI-powered automation. Invoice reconciliation takes 90 hours a year per staff member.
416 minus 90, means you’ve got back 326 billable hours a year.
That’s 6.3 hours a week. Almost a full day a week.
Reclaiming Your Professional Life
So, what does all this newfound time mean for you and your business.
Are you able to add more services, like advisory and consulting packages? Does it mean you can serve your current clients at a higher level?
Can you finally focus on strategic business growth activities?
What about your staff? Do they leave work on time? Have they stopped working late?
Have they more time for training, upskilling? Keeping up with the ever-changing demands on them, and you by HMRC?
Are they happier? If so, they are less likely to walk out on you? To quit!
What About Your Personal Life
I can make 8,657 assumptions of how this can improve your life, outside of your business.
Less stressed, better sleep patterns, more family time? More time for you and what you want to do?
The list is endless.
I would love to hear what the time saving would mean for you, both for your business. But for you personally, reach out and let me know.
AI Invoice Reconciliation Implementation Roadmap
OK, what are the practical steps to get this AI-powered automation up and running.
First, my recommendation is to set up a dedicated automation server.
You’ll end up wanting to layer on more and more automation workflows after this one. Once you see the benefits.
A dedicated automation server also prevents automation costs escalating. It’s a fixed amount each month, so you can budget for this.
A dedicated server also addresses privacy concerns. As well as getting backed up every day and monitored for downtime. One less thing to worry about.
The next step is to integrate the server with the accounting software your business uses. For example, Xero.
You or your clients should use an invoice uploading service. Such as Hubdoc to semi-automate the first part of the process.
Finally, a way of producing the report, storing it somewhere and alerting the business. I’ve found Google Workspace useful for this. Utilising Google Sheet, Drive and Chat.
Implementation Timeline
After initial discussions and a kick-off date. It can take 1-2 weeks to get the first workflows up and running and handed back to you for testing and review.
I’ll be upfront. You need to be available for an hour or so during this time. To help make sure the automation server has access to all the relevant systems.
Once the testing phase is over. Depending on the number of clients you are working with. It can take 1-2 months to get full deployment of all the workflows done and dusted.
A 3-month optimization period after the deployment is good practice too.
This is so any tweaks or edge cases that occur get reviewed to improve the automation.
Also, a word on new clients. As part of the onboarding process. A new workflow may need to be set up.
This will take less than a day. In a most cases a few hours.
Staff Training
You’ll also need to be aware that staff may need training on the new AI invoice reconciliation workflows.
In my experience. And one of my goals when working with a client. Is to make sure any automation brings as little disruption to a process as possible.
But it’s something you need to keep in mind.
Check out this article I’ve written about getting team buy in for automation projects too.
There can be push back, this article lists the 5 biggest fear your staff may have. And how to deal with them.
Conclusion: AI Invoice Reconciliation
So, what is the benefit of implementing an AI invoice reconciliation process?
Well, an 8–10-hour job, which gets done every week. Now takes just over 1 and a half hours to complete.
It frees up between 6-8 hours a week for you and your staff.
Over a year, this gives you back over 300 hundred billable hours per person.
What does that mean for your accountancy business?
Less staff burnout, happier workforce. Better at their job, which means happier customers.
Now to the biggest benefit. Work-life balance.
Not only for you, but also your team.
Can you take more time off? Have more time for those important to you? Even take a few hours a week for yourself!
One thing I haven’t mentioned is the competitive advantage you’ll get.
Happier customers, fewer leaving. They tell more people how amazing you are. More new customers.
It could be a snowball effect.
If you want to see if this type of automation could work for your business. Check out my consulting page. More than happy to run through a demo of how it could work for you.
Or discuss a full AI Invoice Reconciliation implementation for your accountancy practice.
