Automation as self-driving glass box - Aera's Fred Laluyaux on the firm's Cognitive Operating System platform play

Publisher: Diginomica

Author: Chris Middleton

Cognitive software provider Aera Technology has announced the launch of its new Cognitive Operating System. It describes the offering as “world’s first cloud platform for cognitive automation” – the automation of Artificial Intelligence, machine learning, data modelling, and data science functions to help run business operations.

The platform allows employees to build and configure their own digital Skills – automated processes that combine different functions and data sources – use prepackaged Skills, or work with partners on developing Skills for the organization. These might include demand planning in the supply chain via harmonising data from ERP, CRM, sales, marketing, trends, and social feeds, for example, or optimising trade promotions.

The move pitches the company into a big blue space that is also occupied by the likes of Blue Yonder (formerly JDA Software), the UK’s Blue Prism, and others that use a mix of AI and digital employees to automate supply chain operations and others. Aera boasts blue-chip clients such as Johnson & Johnson, Unilever, RB, and Merck. That’s a lot of blue in a hyper-competitive sector that – purely in branding terms – seems more focused on big skies than drilling down into granular detail.

The key theme for Aera is that its new platform enables the ‘Self-Driving Enterprise’, a term that the company has trademarked. The announcement says:

Aera harmonises both internal and external data across the enterprise; applies science (analytics, search, optimisation, modelling, and data science) to derive insights and recommendations; digitises, automates and augments decision processes; and operationalises change by digitising institutional expertise and experience.

What this means in practice is that Aera’s approach to decision-making, automation, and augmentation is one of constant learning and iterative improvement, crawling through data sources and storing memories of its decisions and the employee actions relating to them. A digital brain for the enterprise, claims the company. Laluyaux says:

We've been pioneering the concept of cognitive automation and augmentation with our clients, some of the largest companies in the world. We built a series of engines that meant every deployment was custom made. All of that now is available through an elegant SaaS platform.

The last component is the change necessary, so one of the things we've learned and incorporated into the platform is, how do we enable a large organisation, globally, with a cognitive skill at scale across multiple business units, market units, customer segments, or product segments, and we built some tools for that.

And how do we enable those organisations to actually take their hands off the wheel and allow the system to run in self-driving mode? So we built a series of tools, a cognitive workbench, a cognitive decision board. All of these are being released along with the platform.

So really today’s announcement is everything that we've learned about those things with pioneer accounts for Johnson & Johnson, Unilever, or Merck – how to make cognitive automation work. We're really changing the way work is being done. It was clear to us that we were going to be walking into uncharted territory, so we had to first take everything that we've learned and package it into a platform.

The new offering is available to some clients already, but will be released more widely within “a couple of months”, he says. But the bigger question is should enterprises be self-driving? After all, might a self-driving multinational ultimately devour the planet’s resources in its quest for constant growth, revenue, productivity, and profit? What does he mean by ‘self-driving’ in this context? Laluyaux says:

The area where we learned the most is around process and change. Think about a self-driving car: we use that analogy quite a bit. It's really the process of letting go of the steering wheel, and letting go is not easy because that's your job, you've been working as an operator in a company and you’ve learned to make optimised decisions on an ongoing basis.

To let this digital brain actually make some of the decisions – right back into the transactional systems – and change the way a truck is going to be dispatched on the road or inventory orders are going to be made, we had to learn how to establish some trust. How do you move from people doing the work to actually having the computers do the work, run the forecast, build the promotion plan, and so on, guided by people?

You want to allow the business users to influence the way this ‘brain’ actually works. So again, we had to come up with the tools that allow that interaction to happen very elegantly and efficiently. It's obvious that once you remove the bias from the decisions you get better performance, which is what the digital approach allows us to do. You know, bias is what we do as humans.

Bias issue

The idea that humans introduce biases that technology is best placed to remove is a challenging one. There has been intense debate about bias in the training data for AI systems, for example, which then automate historic problems. But to propose that humans need technology to correct their biases is risky – even if it may be true. After all, who is to say that Aera hasn’t made the same mistakes in system design and market assumptions as others, in partnership with its enterprise co-developers? Who is to say that Merck’s or Unilever’s view of the world is one that should be automated globally?

An emerging need is for organizations to be able to show their workings in an increasingly automatic and AI-informed world, so that customers can see if there was any bias in the decision-making. How easy is it for organizations to expose the decision-making process in Aera to anyone who needs to see it? Lalauyaux says:

It's actually easier, because in the model the software allows me to understand and visualise every step of the way – as in, this is how the recommendation was made. So actually it's much easier than when you're digging into people's minds, so to speak, to get to the same result.

It gets a little tricky when it comes to data science because not everybody understands how predictions, classifications, or optimisation work, but overall our clients are getting much more savvy about it. I would say that the transparency that the digital system provides is completely there. We talk about a glass box, as opposed to a black box.

A lot of the software that has been built over the last 30 years is actually quite opaque, in that you don't really know what's happening inside, what kind of processing is going on, and you have to complement that weakness with very experienced operators who have been working for maybe 10 years in a specific region or planning for specific products. And that can be a worry, because it was a combination of experienced people with opaque software.

With automation you actually need to have a much higher levels of transparency, because you don't necessarily need that experience anymore, old time. However, you need to understand the process in real time. Why is Aera making such a recommendation? So that's the goal by mapping the problem.

He adds:

We talk about Aera being guided by people, which means that the people that were doing a lot of that repetitive work are now sitting behind the machine and making sure that the machine is delivering the right recommendation and enabling the right actions.

The trust comes from it building that interaction between informed users. What the machine delivers is massive scale – it’s 24/7, the ability to measure a vast amount of data and processes at internet scale. That’s much better than any group of people can do, equipped with their own individual and collective tools.

That's the key: building that intimacy between the algorithms and the users, and allowing the system to get better over time.

So how big an issue is the cultural change needed within organizations? He says:

I like the analogy of the Uber or Lyft driver. What's really happening is you're in this car and you are completely driven by a computer system, a system that tells you where to go, what to do, how much money you're going to make. It tells you which road to take to go from A to B, and as an Uber driver you're basically operating based on instructions that are given to you by a computer.

We believe we can do better than that, we can put the human back into more human functions and let the computers do the basic automation. And I think if you look at large organisations, they have a lot of the equivalents of Uber drivers – people who have been driven by systems, who were trying to follow processes and use a bunch of tools, but they've not really been on top of that process.

So I feel like it's a liberation of process that we're going to see, as opposed to further alienation. I think that the alienation is already there. We just fail to acknowledge it.

My take

Make way for the self-driving glass box. This is all well and good, and Aera’s cloud platform was the obvious next step in presenting its tools to enterprise users. But the cultural change necessary for humans to take their hands off the wheel is bigger than many business leaders realise, especially as most consumers are telling companies to put their hands back onto the wheel and simply drive more responsibly. Can Aera help them do that? To answer that, you may have to look into the glass box.

See Aera in action.

Meet Aera