Seeing things clearly

A few days ago, the Financial Times’ Laura Noonan published an article, ‘Commerzbank sets AI to work writing analyst reports’, reporting on our work with the German banking giant to use Natual Language Generation (NLG) technology to write basic analyst notes.

As one investment bank head told Noonan, “There’s definitely work that can be done, [and] parts of the [research] process that can be enhanced by algorithms and AI tools.”


Where we think the article is slightly opaque is about the true effect of automation on human analysts. Just because automated content will take some of the work from analysts, it does not mean that it will take their jobs. Robots may be quicker and more efficient than their flesh-and-blood counterparts, but they still lack fundamental skills that only humans have. This is something that we have touched on repeatedly, especially here, here, and here.


In each of those articles, written over the last year, we outlined what robots could and could not do. Our view, at each step of the way, has been that automation will free people up to do the more-creative tasks that only humans can do. Or, as I wrote back in August, “Ultimately, poetry-generating machines fail. Not because they fail to produce poems but because they fail to produce anything outside the narrow lines it is given. We can teach a machine to do something, but they’re not yet at the point where the machine can learn it by themselves.”

The limits of a robot’s capabilities should be the beginning of where a good analyst brings quality. Robots are limited by the dataset they work from, meaning that they are blind to outside influences; cannot always find connections or give context; and lack the ability to make complex comparisons or be able to parse the bigger picture.

On the other hand, robots are faster than humans; do not get tired, hungry, or thirsty; work twenty-four seven; and do not make mistakes in calculation. However, they are only as good as the datasets they are fed and the parameters they are designed along, which they yet cannot generate or form themselves (making them reliant on humans). Noonan herself touched on this in April in her longform article ‘AI in Banking: The Reality behind the Hype’ (we responded to that article, too.)


Noonan’s report generated a number of comments and responses. There were a few interesting thoughts among them where those writing recognised the value that NLG content could bring to their industry. As one commenter wrote, “AI should be a tool to augment and improve research. If you make it purely about cost reduction, that’s a road to poor quality.”

They are right. Companies should not be looking to NLG and AI in order to reduce costs (although that can be a welcome benefit). The real value is in supporting quality by handling the mundane tasks that make up the quantity of a workload. That is why the first sentence of that comment is the most true: AI should augment and improve already-existing practises.

Perhaps the most-salient response to the article came from someone using the name ‘Koba’, who wrote: “Much talk of ‘replacing’ analyst research but such linear thinking rarely produces results. We can see much value in using AI to aggregate and crunch earnings reports. Machines will definitely have an edge over humans in collating and comparing vast numbers of earnings reports in real time, extracting trends and outliers for future analysis. It will be some time however before machines can parse earnings calls and form a view of market psychology. Hopefully, this technology will automate that which should be automated and allow the expensive human analyst to concentrate on judgment calls on an expanded universe of stocks.”

‘Koba’ is right. An NLG product should not be developed with the aim of replacing an analyst. It should instead buttress the work of that analyst.

It seems that many fear automation because they overestimate the current capabilities of NLG and fear that such projects are intended as replacements for humans. This is very much not the case. A good parallel can be drawn with the wearing of spectacles. Nobody would pretend that wearing glasses is a replacement for having eyes.

The truth is that some organisations may seek out NLG as a way to reduce their workforce. But that would be a shortsighted and misguided motivation, a surefire way to harvest quick, short-term profits but sow the seeds of failure. And any company that seeks to replace employees this way will seek to replace those employees by any method. It won’t be the NLG that kills jobs, but how it is developed and deployed.



About Retresco

Founded in Berlin in 2008, Retresco has become one of the leading companies in the field of natural language processing (NLP) and machine learning. Retresco develops semantic applications in the areas of content classification, recommendation, as well as highly innovative technology for natural language generation (NLG). Through nearly a decade of deep industry experience, Retresco helps its clients accelerate digital transformation, increase operational efficiencies, and enhance customer engagement.