The accounting expertise disaster is altering how companies deal with their funds. With the US market needing 340,000 new accountants and 45% of companies struggling to draw expertise, extra firms are turning to specialised expense administration suppliers to deal with their accounting operations.
However here is the catch: These service suppliers face the identical expertise scarcity.
We just lately labored with one such supplier who processed over 50,000 payments and invoices month-to-month for his or her shoppers. As extra firms got here to them for assist, they bumped into the identical impediment: they could not rent sufficient individuals to maintain up with the high-volume workload. It obtained to a degree the place they needed to flip down new shoppers.
Let me take you thru how the service supplier applied automated bill processing and located a option to bulk-process payments and invoices with out continuously including extra workers.
What occurs when rising bill and invoice quantity breaks doc processing
The expense administration supplier in query makes a speciality of serving to organizations optimize their working expenditure. As a Enterprise Course of Outsourcing (BPO) companion, they deal with all the pieces from utility and telecom bills to IT prices—processing payments, validating fees, and offering detailed analytics. They assist AP groups in healthcare, manufacturing, retail, and authorities organizations cut back their workload whereas delivering price financial savings by way of higher expense administration.
Nevertheless, as this supplier’s enterprise grew quickly, dealing with near 50,000 utility payments and invoices month-to-month grew to become greater than only a processing problem. It uncovered basic operational constraints that threatened their means to scale and ship worth to shoppers.
Here is what occurred:
1. Handbook processing grew to become a bottleneck: Their crew relied on handbook knowledge entry and will solely course of about 100 invoices per individual per day. It created a rising backlog and restricted their capability.
2. Expertise scarcity worsened issues: Discovering and retaining certified in-house accounting analysts to supervise the method and guarantee accuracy grew to become more and more tough and costly. This made it tougher to scale their crew and meet rising calls for.
3. Consumer service suffered in consequence: Processing inaccuracies and inefficiencies meant delays, errors, and, in the end, sad shoppers. Additionally they struggled to satisfy consumer deadlines, resulting in potential penalties.
4. Revenue margins have been squeezed: Elevated labor prices and SLA penalties reduce into their profitability, making it tougher to put money into development and innovation.
5. Development alternatives have been restricted: They could not effectively onboard new shoppers or tackle new enterprise alternatives. It put them on the again foot towards opponents with higher techniques.
They wanted to take care of a big crew of knowledge entry clerks always to key within the sheer quantity of payments and invoices. Whereas they have been capable of outsource a few of it, the inherent limitations of handbook processing remained.
The tipping level was when the service supplier began turning down new enterprise as a result of their operations couldn’t scale to serve them.
Why bulk bill processing grew to become difficult
Check out a typical utility invoice, it would look like a easy doc to course of. However it might shortly turn out to be a headache if you’re processing hundreds of them in bulk every month. Extra so if you’re aiming for the accuracy and element wanted for efficient expense administration.
Every bill accommodates important knowledge factors that have to be extracted and validated: account numbers, service addresses, meter readings, utilization knowledge, cost descriptions, and cost phrases.
Here is a fast glimpse into how difficult the utility invoice processing workflow was:
- Each utility invoice is totally different
- Tons of of utility firms, every with distinctive bill codecs—some even utilizing 7-8 totally different layouts
- Some utility firms ship each day payments, others month-to-month
- Single invoices can run from 1 to 372+ pages
- Every format wants particular dealing with
- Billing info is not simple
- A number of meter readings for various companies (electrical energy, fuel, water) may very well be scattered throughout the doc
- Service addresses for multi-location shoppers is perhaps unfold throughout pages, or buried inside tables
- Cost descriptions lacked consistency (“Buyer Cost” in a single invoice and “Cust. Chg.” in one other)
- Equivalent fees may seem a number of instances on a single bill, requiring handbook revisions
- Accuracy necessities are strict
- Meter readings should match their corresponding fees
- Information wants validation earlier than system import
- Errors may result in incorrect consumer studies, flawed monetary insights, and missed optimization alternatives
- Processing delays result in SLA penalties
These complexities imply handbook processing or primary OCR options, which frequently battle with inconsistent layouts and complicated knowledge relationships, wouldn’t be sufficient. They wanted an answer that would not solely deal with excessive quantity but additionally adapt to the intricate particulars and variations of their shoppers’ utility payments.
Reworking high-volume bill processing with AI and automation
The expense administration supplier knew they wanted to automate bill processing earlier than development grew to become unimaginable. Nevertheless, with 50,000+ month-to-month invoices, complicated validation necessities, and strict consumer SLAs, they could not afford any drops in accuracy or service high quality through the transition.
That is when their CFO found Nanonets. Our one-shot studying, customizable validation flows, hands-on assist for complicated vendor setups, and talent to deal with multi-page tables satisfied them that we may meet their particular wants:
- Distinctive doc layouts
- Advanced knowledge extraction wants like a number of meter readings and repair addresses
- Excessive-volume processing with constant accuracy
- Validation necessities for his or her system import
Working with the supplier, we took a scientific method to automation. Moderately than disrupting their complete workflow without delay, we began with a vendor-by-vendor method to validate accuracy and construct confidence.
At this time, their automated workflow processes a good portion of their 50,000 month-to-month payments and invoices. Right here’s the way it works:
1. Doc consumption
- Their crew collects invoices all through the day by way of a number of channels (mail, consumer emails, utility firm portals, and bodily invoices)
- On the finish of every workday, they place all collected invoices into a delegated SharePoint folder
- The SharePoint folder construction can go as much as 3 ranges deep (e.g., fundamental folder → vendor folder → date folder)
- Nanonets checks the SharePoint folders each 3-5 minutes for brand spanking new recordsdata
- When new recordsdata are detected, Nanonets routinely imports them
Be aware: Backup strategies embrace direct add by way of the Nanonets interface, e mail forwarding to a devoted mannequin e mail deal with, and different cloud storage choices (Google Drive, Dropbox)
2. Preliminary processing
- As soon as imported, Nanonets’ AI mannequin begins extraction instantly
- System identifies and processes:
- A number of meter readings as desk constructions
- Advanced cost descriptions
- Service addresses throughout pages
- Cost phrases and dates
- Constructed-in validation checks the payments for:
- Discipline completeness
- Date format accuracy
- Quantity validity
- Required info presence
Be aware: Their crew adopted an end-of-day batch add method, permitting our system to course of all the pieces in a single day. This ensures all paperwork uploaded through the day are processed and prepared for assessment the following morning.
3. Error dealing with and high quality management
Error dealing with entails automated checks, adopted by handbook opinions for corrections.
A number of the automated checks finished by Nanonets:
- Flags lacking fields
- Identifies mismatched meter readings
- Catches incorrect cost descriptions
- Tracks accuracy charges by vendor
The processing crew then opinions error logs and output each day for:
- Cost descriptions within the fallacious columns
- Service addresses throughout a number of places
- Meter readings matching their fees appropriately
When points are discovered, the crew investigates the basis trigger, corrects extraction errors straight within the Nanonets interface, and updates the AI Mannequin’s coaching recordsdata if wanted to enhance future accuracy. This steady suggestions loop helps keep excessive accuracy charges throughout all distributors.
Be aware: Every day error reporting was applied to stop points from bleeding by way of your complete week.
4. Submit-processing and output
As soon as paperwork move validation:
- System applies customized formatting guidelines to extracted knowledge
- Standardizes cost descriptions throughout distributors
- Information is exported in customized CSV/Excel format
- Information are organized by vendor and date
- Information is imported into the service supplier’s billing system
Be aware: The crew processes 15,000-20,000 pages month-to-month this manner, which may scale to 30,000+ pages.
5. Steady enchancment
We work with the service supplier to take care of and enhance processing high quality. Their crew identifies areas for enchancment and offers suggestions, whereas we constantly improve the AI mannequin’s accuracy. This entails:
- Common mannequin updates based mostly on their processing crew’s suggestions
- Centered enhancements in complicated areas like cost descriptions
- Gradual onboarding of recent distributors to make sure steady efficiency
- Utilizing our analytics dashboard to watch and optimize efficiency
This collaborative method helps keep excessive accuracy charges whereas steadily increasing processing capabilities.
The affect of automation on bulk bill processing
Inside three months of implementing Nanonets, the supplier noticed vital outcomes. For starters, they managed to automate processing for ~12,000 invoices monthly. This highlights the system’s means to deal with their high-volume wants.
For utility invoices from identified distributors, the AI mannequin achieved processing speeds of 173-174 invoices in simply 5 minutes — a process that may’ve required hours of handbook knowledge entry.
However the affect went past simply velocity.
Processing effectivity
- Nanonets can now course of 46 out of fifty distributors shared by the client
- Just one consultant file required per vendor for AI mannequin coaching
- Automated doc consumption by way of SharePoint integration
- Bulk in a single day processing of 500 invoices each day
Operational enhancements
- Employees free of handbook knowledge entry to give attention to higher-value duties like consumer reporting and evaluation
- Quicker turnaround on consumer deliverables and assembly SLAs
- Extra constant processing high quality, with fewer errors and handbook corrections
- Higher dealing with of complicated invoices with a number of meter readings and repair addresses
The trail ahead
The implementation remains to be within the early levels. With solely ~20% of the supplier’s complete invoice quantity at present automated, there’s vital room for progress.
Their subsequent steps give attention to:
Increasing automation protection
- Processing telecom invoices (estimated 10,000 month-to-month)
- Scaling to deal with ~30,000 payments and invoices month-to-month
- Persevering with to onboard new invoice codecs from totally different utility firms effectively
- Constructing on their profitable utility bill automation
Technical integration
- Implementing API integration by summer time 2025 to allow direct knowledge circulate between Nanonets and their expense administration answer
- Automating error corrections the place doable to cut back handbook assessment time
- Constructing a extra strong integration between Nanonets and their in-house options