Let's say it as it is, administrative functions are still at the stone age of digitisation. Finance, HR, Procurement departments rely on laborious and manual processes including tons of paperworks are still waiting for the digital age.
The (non-)digitisation of administrative functions has a direct impact on a company's competitiveness
There are many reasons to this lack of productivity including the fact that CEOs and CFOs prefer to focus on increasing revenue, direct margin and top-line.
But one of the biggest hurdle that makes those functions complex to automate is the volume, the complexity, the heterogeneity of the documents and data to be processed. In addition, administrative documents and data are often disseminated in various locations and departments and the back-office processes are most of the time siloed.
Just thinking about a simple document like the supplier invoice gives headaches. Let's take the example of a 1B$ multi-national company, that has 1200 suppliers and receives 100k invoices per year in the 30 countries its operate. In the best case most of those invoices are digitised (non-paper) but in average 80% is non structured data (image pdf). It costs in average $4 to process each invoice manually, 6 mn to enter it in the centralised accounting system and an additional 2 mn to allocate it to the correct spend category in the global procurement system. In terms of headcount, the process implicates almost 100 people from 3 departments, procurement, accounting and finance control, just to register this basic document [^1].
This has a direct impact of the enterprise's productivity and competitiveness.
Artificial Intelligence enables fast digitisation and automation at low cost
As a reminder, the expected benefits of automation are 3 fold:
- decrease of costs, mostly through headcount reduction
- increase of productivity, enabling humans to focus on value added tasks
- increase of quality by removing errors, ensuring consistency in an comprehensive way.
Many efforts and investments have been made for years to remove the burden of paperwork and manual processes. Unfortunately, this has often been limited to the digitisation of paper documents, the outsourcing to low cost countries, the deployment of people intensive workflow tools and an overuse of spreadsheets. Hardly meeting any of the automation objectives...
It is the combination of Artificial Intelligence (AI) and Big Data that changed the game. Once being trained thanks to huge sets of data, AI algorithms, so-called Machine Learning algorithms, can help operatives in many manual and tedious tasks. AI algorithms are capable of being autonomous in the following fields; data ingestion, manipulation of complex, non formatted and unstructured data, classification and organisation of data, prediction, decision making, action automation
For example, AI algorithms can perform an end-to-end automation of the supplier invoice management from the pdf invoice entry in the information systems to the spend classification for the procurement teams, the payment to the suppliers and the provisioning of the general ledger for the accounting team. In the example of our 1B$ multi-national company, this means a potential saving of millions of $, full visibility on the spend and the guarantee of paying the negotiated prices to suppliers.
There are many other use-cases where IA and automation provide short term benefits and strong ROI in the back-office functions. For example:
- supplier invoice entry: Deep Learning algorithms to read and understand Invoices like humans and replace the limited OCR technologies
- automated accounting: Machine Learning systems that automate the entry in the general ledger
- invoice compliance: automatically identify duplication, pricing, VAT & volume errors in supplier invoices for example in the long tail spend
- automated spend classification: classify automatically and continuously the millions of spend items into a centralised and dynamic enterprise referral
- category cost optimizer: for greater sourcing management and increased savings
- automated payroll control: guaranteeing complete compliance with local labour legislation without the need for time-consuming manual checks
- **absenteeism prediction**: supporting manager identifying workers at risk
The raise of the Enterprise automation platforms
In order to ease the introduction of AI and automation in the Enterprise administrative functions and to avoid ad-hoc non consistent projects, the algorithms need to be packaged as software. Those software need to have following characteristics in order to be cost effective, future proof and to maximise the automation benefits :
- Cloud based to centralise the data and the processes through the whole organisation and to mutualise the algorithm training sets
- Flexible to ingest any structured or un-structured data, regardless of its location or format
- Open to connect seamlessly and at no cost the enterprise legacy systems such as ERP, reporting tools, or usage databases but also systems from the enterprise eco-systems like suppliers
- Powered by Machine Learning, to be trained by operatives for the set-up and for continuous improvements
- Automated to be able to replace simple manual tasks but also more complex decision making
- Enriched with business expertise to deliver short term ROI and maximize automation
Keeping humans at the heart of the Enterprise
One day in the future, machine and robots will be capable of learning by themselves and discover what to automate without being taught by humans. But this is not the case today, like children, AI based software needs to be trained by experts. In addition, many decisions and some tasks still need to be performed by enterprise employees.
This is why Entreprise automation platforms today need to be tailored for each department as vertical applications. In order to:
- ease the teaching of algorithm by operational people, who need to see short term ROI from the tool
- to integrate seamlessly non-automated processes in order to offer business continuity and avoid loss of quality
- onboard business expertise whenever it is possible to maximise benefits
- provide an optimal user experience
Deployed as stand-alone SaaS web services or as add-ons of legacy systems such as e-procurement solutions, accounting sofware or ERPs, those vertical applications assist the administrative function's employees in their daily work by automating their manual tedious tasks, recommending actions and by acting as a team of hyper-vigilant auditors working around the clock.
Therefore buyers, accountants, finance controllers or payroll controllers not only augment their productivity but also increase their value in the organization by focusing on more strategic and game-changing activities.
For the benefit of the enterprise and for an enhanced personal satisfaction.
[^1] Dhatim data