Artificial intelligence is in the midst of its next wave of innovation - and avvaneo is not to be outdone

Our History of AI Development

avvaneo has been focused on AI for a long time. Early on, we saw various opportunities to integrate AI into financial processes,starting with work on chatbots and intelligent voice control. At the beginning of 2019, we began developing an AI solution with tangible customer benefits that takes automation to the next level.

The result was avvaneo Expedite. Our customers have been using this solution since 2020, registering significantgains in efficiency immediately after implementation; these gains continue toincrease with each round of training.

Today, our entire product team is working on AI. avvaneo is now  an AI service provider for our own software products as well as providing automated document capture and processing in the SAP area.

The primary problem for an AI solution is often the quality and quantity of the data. Accounting departments have enough historical data, but they are often structured very differently. Nevertheless, avvaneo was able to reach its goal relatively quickly for two reasons: First, we have twenty years of expert knowledge on modern accounting systems, their document references, and their process steps. Secondly, we are no strangers to dealing with heterogeneously formatted documents and proprietary structured data . We therefore knew immediately where a machine learning module would surpass manual programming.

What data do we use to train and test our models?

Artificial intelligence detects recurring patterns, searching for  similarities in the largest possible amounts of structured, semi-structured and unstructured data. Our training sessions serve to discover patterns even in heterogeneous, erroneous or "rule-breaking" documents. For this learning process, historical data is ideal, i.e. documents such as purchase orders, delivery bills and invoices for which the valid solutions are known.

The AI then tests the learned pattern recognition with unknown data, ideally from the company in which the models are to operate. In this way, company-typical regulations, contractual agreements, or internal rules are learned right away.

Gradual self-improvement is the greatest advantage of machine learning over manual programming. So we optimize our AI systems at regular intervals based on new data sets. The beauty of this is that the learning takes place behind the scenes, while the AI continues to do its work automatically.

2023: What process points can avvaneo Expedite improve today?

The financial processes around receiving invoices can add up to a hefty workload in many medium and large companies; every non-automated detail  can therefore add up to a significant cost factor. avvaneo Expedite can significantly increase automation in these (and other) tasks today:

• 1. Determine and verify the account assignment of documents

• 2. Forecast processing times to optimize discounts and meet payment deadlines

• 3. Send correct workflows for review and approval

• 4. Highlight suspected duplicates to prevent double payments

• 5. Detect anomalies to prevent errors, manage and improve audit processes, or detect fraud

• 6. Assign items found on invoices and purchase orders

• 7. Determine tax codes, partner bank types, state central bank indicators, etc.

Automatic account assignment and correct determination of workflow recipients were the first important features that achieved a new level of automation with our AI module. In the meantime, other features have been added; for example, automatic assignment of items between different documents (such as purchase orders and invoices), or the accurate determination of tax codes.

What's in it for you?

Significantly increased productivity and efficiency. Artificial intelligence measurably reduces the number of documents that require manual processing. As a result, avvaneo Expedite achieves automation rates of 40% to 70% almost from a standing start.

And that's not all: avvaneo Expedite's analytical expertise can also stop the automatic passage of questionable documents with anomaly and duplicate detection and alert employees to the questionable spots.

Our implementation is not yet automated, but will be soon

In some cases, implementation is completed within a month; in others, it takes up to three (although the pure project time is significantly less). Our system has been developed to introduce AI into the company faster, cheaper and much safer than would be the case with in-house development. Implementation takes place in four steps:

• 1. Pilot project based on real data from the company: It analyzes the optimization potential and is created in the existing system. The most important advantage of the pilot is that it can foresee concrete results and an optimal approach at a very early stage.

• 2. Implementation project with the results from the pilot: Thanks to the findings from the pilot and our deep knowledge of SAP systems, the integration of our solution can be implemented quickly and painlessly in most cases.

• 3. Launch: When avvaneo Expedite is productive, employees quickly acclimate to the new possibilities because the improvements run mainly unnoticed behind the scenes without changing the processing interface much.

• 4. Continuous optimization: Implementing AI means starting a continuous optimization process. The implemented models will be retrained at regular intervals without interrupting the ongoing operation. Over time, the module captures the typical invoicing patterns of the company better until the models reach their maximum potential.

Our Goal: Fully Automated Accounting

The GPT disruption has quite suddenly shown the unlimited possibilities of artificial intelligence. This also applies to our AI development. We already have our sights set on the goal of a fully automated accounting system. Our AI solutions will then no longer only support individual parts of the process but will handle the posting of incoming invoices in full. This could even include automated communication with the invoicing parties or company employees.

You can imagine this development like the development of cars. First, more and more functions are supported by assistance systems, then taken over by them until, finally, fully autonomous driving is a reality. Similarly, we are also on the way to more semi-autonomous intermediate stages until finally fully autonomous accounting will be possible.

We look forward to taking the next steps with you. That it will happen is no longer in question.