There are differing opinions and discussions surrounding the potential impact of artificial intelligence (AI) in the financial sector. Looking to reassure firms that AI can be used ethically and safely by organisations, we spoke to DeepOpinion.
Earlier this year, BT, the telecoms giant, made headlines by announcing plans to potentially replace 55,000 roles with AI, sparking a debate across various industries, including fintech, about the implications of AI’s emergence. While concerns have arisen, it’s worth noting that AI can be used ethically and effectively, often functioning best as a complement to human workers rather than a replacement’, according to DeepOpinion.
DeepOpinion is powered by large language models (LLMs) that automates the operational complexity of training human-level AI models. It enables teams to easily create intelligent automations for text and document processes, without code.
The initiative leverages deep learning where it can connect to any text or document source via API or drag files and folders from the client computer, then quickly label the data to create a machine learning dataset.
To learn more about how AI can be used effectively, we spoke to Ahmed Al-Ali, co-founder and region director, DeepOpinion.
How does DeepOpinion address ethical concerns related to the use of AI, especially in industries where decision making is crucial?
This is a crucial topic for enterprises across a variety of sectors. AI ethics and AI safety impact different players in different ways. We focus on processing documents and providing certain controls that ensure the quality of the output of AI is compliant with what organisations would expect.
For example, we have set up our technology in a way that when AI spots an anomaly, it can notify a human to review its proposed solution. When extracting information from an invoice, it can say for example, ‘I’m not sure this is the customer’s address. Can you have a look and check it?’
Our machine learning team also employs certain quality tests to ensure that the performance and the output of the AI model is up to standard. Furthermore, we have a deep focus on research, to identify ways in which we can invite users to understand how AI comes up with certain decisions.
Can you provide examples of industries or applications where human level AI is making the most substantial impact? And how does DeepOpinion contribute to these areas?
Generally, AI can act in a couple of ways. The first is ensuring autonomy. We can see examples of this in self-driving cars, and manufacturing robotics. However, we also see applications where a person is needed in combination with AI to complete a task. This is typically the case where you don’t want the AI to go on autopilot, but rather, you want it to do some background tasks and you complement its output to reach the quality that is desired.
You get a balance between highly reliable output as well as speed and efficiency. The area which we focus on is mainly in automating processes within organisations. The communication of organisations with the external world tends to be in an unstructured data format. For example, you communicate with your customers and suppliers, in emails, chats, documents, and so on, and typically there are big teams within an organisation that try to make sense of this.
This puts a massive overhead on those organisations to complete repetitive non-fulfilling tasks. With AI you can automate a lot of these processes and make them run on autopilot. Meanwhile, humans can still be called for handling exceptions, but at the same time, have freed up time to focus on more meaningful things and this is the areas that that we are focusing on.
What are some specific use cases?
Typically, the industries that we focus on are document heavy industries. These often tend to be manufacturing and financial services firms, which can be further split into banking and insurance firms, as well as government institutions.
These industries have plenty of documents that they need to handle. We’ve managed to deliver a service that has helped them automate those processes to an incomparably good quality.
We consider ourselves to have world leading performance when it comes to deploying large language and AI models to tackle the processing of documents. Some success stories that I can think of include our work with Siemens. We are helping them automate their delivery notes within their manufacturing plants; automatically processing and adding them to their ERP systems without human intervention.
As another example, we’re helping one of the largest banks in Austria to handle their account payables and process their invoices. We’re also helping a stock and crypto trading platform to automate their customer onboarding process, where they do customer document checks and KYC as well.
How can you reassure someone that AI is a good thing?
From a business standpoint, there are multiple answers to this question. Firstly, if you’re not leveraging AI in your business, you’re positioning your company to be less competitive. In our current landscape, it is unlikely that your competition will have the same concerns, and they will leverage the latest technology, gaining an advantage over you. No firm wants to be left behind.
Same thing applies to employees on a personal level. Research from the World Economic Forum has said that there are very few repetitive jobs that might be replaced. However, this won’t happen now. It will take a while – at least five years to be adopted by organisations. Furthermore, it found that in more than 60 per cent of cases, AI was able to help with some tasks but was not capable of doing the whole job. Therefore, it will make employees more productive instead of replacing them.
AI won’t necessarily take your job. But someone who’s using AI, leveraging it and understands it will. It’s better to upskill ourselves to understand the technology and to leverage it.
Both customers and shareholders expect you to leverage AI for a better quality, top line and bottom line. Therefore, for those that are concerned, I would say try to understand this technology. Understand how it impacts you, how it can make you do a better job and ultimately how it can serve your customers better.
Having said that, implementing the right mechanism to be aware and control AI risks is also important. This comes in the form of compliance, adopting an AI ethics and governance framework.
What role does user feedback play in refining and improving DeepOpinion’s AI solutions?
It plays a massive part. We didn’t start assuming that we know all the answers. Yes, we have our expertise in AI, but we want to make a practical useful solution for our customers. We work very close with our customers, and collect feedback on multiple levels. We do so from surveys, customer interviews or simply through usage of our product. We get feedback to refine our service and make it meaningful.
In a more applied process, we also involve our innovator and early adopter customers to co-shape our product roadmap to ensure that we deliver a product that customers want to use and benefit from.
We employed this process when developing our new product line, the “AI CoWorker”. We designed it to solve more complex use cases related to handling documents, once that requires higher-level reasoning and synthesis of information just like humans do. In co-developing this with our customers we uncovered deep insights into their needs and currently shaping the product to get the job done.
The AI Coworker helps employees to have a colleague that they can delegate tasks to. for example, a recruiter can delegate CV screening against job requirements to it or a legal counsel can delegate reviewing NDAs to it and much more.
In this regard, AI becomes more interactive helps create a future where the workforce of an organisation is a hybrid mix between people and AI. In turn, this will help them to achieve more, be more productive and do things that neither can do individually.
Could you share some success stories where clients experienced significant efficiency gains?
I can give you a few examples. One of our financial services clients used to receive onboarding requests through a customer service ticket. An agent would then look at this request and review the documents, store these documents, create an account for the user, onboard them and then send an email to say, ‘now you have an account with us’.
With AI, we’ve managed to tackle a few of these pain points. Now, whenever they receive a customer service ticket, our AI platform categorise and route this ticket by understanding, the type of request and whether human intervention is required for validation. In the case of onboarding, the solution find the person’s ID and their bank statement, extract the information, and match that it is the same person on the passport supplied to automatically complete the KYC and verification process.
If there are any anomalies or anything that needs to be flagged, then it will route them to a human. If not, then it will automatically complete the task on behalf of the team. It can then potentially keep it for a final regulatory check where this is needed depending on the process. We did this for that specific customer, but also in a variety of other places too.
In the manufacturing industry, one of our clients receives millions of documents. With AI they just need to push them through a scanner or take a photo of them and get them automatically processed.
Final thoughts
We are having a journey of a lifetime with my co-founders Stefan Engl and Stefan Ramershoven. Our mission is what we wakes us up every morning which is to eliminate repetitive tasks and reshape the future of work with AI to give humans a more fulfilling contribution at their work. We are also lucky to have a phenomenal team that shares the vision and commitment to make this a reality.
We are just witnessing the first scene of how AI unfolds and contributes to the prosperity of the economy just like how other general purpose technologies did; electricity, the internet, and others. Our product roadmap is exciting and will expand the value we deliver to our customers. We are very much looking forward to continue this journey.
Francis Bignell
