Without a business perspective, ‘deep learning’ AI has a goldfish memory

There are a plethora of start-ups whose AI-based propositions rely on pattern recognition which, while excellent, is often just the tip of the iceberg.

Far from being unintelligent robotic process automation tools, these unicorn tools use AI to recognize and make sense of documents and their content. Thanks to image and text recognition, they can deduce that incoming PDFs or paper documents are, for example, invoices, and thus make an experienced estimation (based on an algorithm or a score) of the individual parts of the content which constitute the name of the company, the tax identifier, or the value to be paid. They can then partially process this information in a semi-structured way (by feeding a database, ticking a box or triggering the next step of a workflow), which notably relieves overloaded teams.

However, while the pattern recognition is excellent, it’s only the tip of the iceberg. True transformation only happens when it can be integrated into the overall business picture, enabling reliable decisions and streamlining prolonged treatments. Without this integration and broader contextualization, even the smartest order or invoice automation solution (or whatever the immediate use case is) will see its potential limited.

Every time the system encounters a new document, it will start the whole process over from the beginning, like a goldfish with no memory.

AI without enterprise context: one-time discoveries and non-actionable insights

In this first scenario, if the AI-enabled automation tool is extremely “intelligent”, it does not accumulate actionable and reusable information about the company or its history as a supplier – or conversely, on his purchasing habits as a customer, nor on his payment history, nor on his degree of satisfaction. It does not provide a 360 degree view of the customer. In addition, unless the installation directly feeds the ERP and financial systems, the last order or the last payment of the company in question will not be accepted more quickly, because its account has not been recognized by the system and even its transaction cannot be identified as “non-fraudulent”.

The AI ​​intelligence capability does not link, for example, between orders and customer history or between invoices and purchase orders. The potential benefits that appear in the form of an improved experience for the company are therefore not as “transformative” as they could be.

The essential next step for enterprise AI is therefore to combine AI and advanced document management best practices with synthesized institutional knowledge, to which it contributes. Thus, across an entire platform and multiple applications, the AI-enhanced enterprise knowledge bank is enriched by every document that traverses it. Memorized in this way, each of these documents contributes to reinforcing the learning of the company, and, consequently, to optimizing its overall vision.

Transforming patterns into strategic interventions

With this powerful interweaving of pattern recognition and contextual AI, the possibilities for transforming business processes are multiplied. Now, once an incoming document has been identified, the AI-enabled content services platform knows not only what it is and what the information within it means, but also what to to do next, namely, which systems need to be updated and which actions can be confidently triggered.

If the document is an invoice sent by a supplier recognized by the system, and the amount corresponds to the purchase order or what was expected, it is possible that long processes (sequential approvals) are, for example, now bypassed which makes it possible to pay a reliable supplier more quickly.

As orders accumulate for a newly added customer, discounts and other incentives may be invoked to reward customer loyalty or, if applicable, to compensate for a recent poor experience. In the case of a human resources or employee-focused scenario, this may involve building a picture, based on sick leave, vacation bookings, access to records or appointment requests. you by worried, anxious or disgruntled employees. This allows line managers to intervene in a timely manner by conducting a review, revising the training plan, increasing salary or granting a promotion in order to prevent valuable talent from leaving the company.

Where apps with “360° visibility” fail

Despite the promises of holistic intelligence so sought after by the myriad of 360° CRM/SCM/ERP/HCM applications, the effectiveness of these systems relies on the information that teams proactively provide to them. Simply using these systems therefore does not guarantee that a sales team will know when they are wasting their time (because such and such a customer is a bad payer), or that a related contract is in progress with the legal team. , or that the personnel of this company have pending requests with the help desk.

By adding advanced AI-powered document receipt and content management, a capability that goes beyond a single application or service, this overview will be automatically enriched continuously, that will allow for better decisions and faster interventions.

Future-Proofing AI: Beware of Single-Use AI

There is another important reason why AI potential should be decoupled from single-use applications: sustainability. A software application marketed “with built-in AI” is, by definition, already outdated. Technology is changing so rapidly that any capability built in today will have a very limited lifespan.

When a software application has the “Built-in AI” function, it means that these companies have bet on a specific AI framework. This can be Google TensorFlow, Microsoft Azure Cognitive Services, a Python-based framework, or specific capabilities for pattern recognition or natural language processing (BERT, ERNIE, etc.). Given how quickly things change and how disruptive cutting-edge technologies can be, it is therefore risky to jump into just one AI framework.

Rather than sticking to a single set of capabilities, companies should instead opt for an “open” content architecture, capable of supporting any combination of current and future AI options on a “composable” as needs change and technology continues to evolve.

Pushing the boundaries of AI

As companies move beyond their AI ambitions of shallow RPA-based automation (which emphasizes capturing/copying and pasting content rather than understanding and modeling of its semantics), it is important that they focus above all on inter-company integration and contextualization. This is what will allow them to fully exploit the potential of AI and machine learning, and deliver that elusive “single view” of customers, suppliers, employees, products, or any other strategic objective.

Any kind of evolution, whether in the global economy, in the reinvention of the workplace, and in the world of technology, is a perfect opportunity for AI to play a larger and more deeply integrated role in a company and its processes, increasing and accelerating the essential daily work that the “human” teams strive to master.

It is in this context that the use of composable, integrated and contextual AI will drive the next waves of process automation, more ambitious and with endless possibilities.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *