The Modern Business Dilemma
Artificial Intelligence - Death by a thousand cuts or go big
Co-Founder, CEO & CTO
Marc Andreesson famously said that "software is eating the world" and never has that been more true. AI is set to add $15.7 trillion to the global economy by 2030 according to PwC and the vast majority of this is expected to come from business automation. McKinsey estimates that 50% of all human labour can be automated with current technology.
Automating business is nothing new, most businesses that are thriving these days are the ones that embraced online, mobile and digital, commonly known as "Web 2.0".
What makes AI Automation different is that where mobile and web were about how you interact with your customers and your employees, AI Automation is about how you automate the core functions of your business.
The primary successful pattern that rose out of Web 2.0 adoption was to find the collection of SaaS products that suit your needs and stitch them together. This has led to super specialism and billion dollar SaaS companies springing up overnight that do one thing really well. The problem is if you take the same approach to automating your core business, then what is left? A brand? This is death by a thousand cuts.
This may sound a little dramatic, so let's take a real world example of a high street bank. While they do a lot of things we're going to hone in on providing loans. According to the 2019 Digital Lending Review only 52% have a fully automated lending process for consumers and 11% for commercial lending.
The manual process pre Web 2.0 or large loans like mortgage:
- Deal with an agent that gives you details of the required documents and information
- You ask any questions about requirements
- The agent answers your questions
- You submit your documents
- Your details get routed to a Know Your Client team, that need to verify that you are not affiliated with restricted countries, organisations, etc.
- Then you get passed to the lending team, who check if you meet certain criteria. If you're asking for enough money you might even get a more flexible decision process than does the application check all these boxes. That process has layers of review and sign-off.
- If the loan is approved, you're back to the agent that sends you the document to sign, answers your questions and actions the loan.
If you're wealthy enough for personal banking, this is quite a nice experience but the reality for most people is hours of elevator music as you wait to be connected. Constantly speaking to a new person who doesn't know all the details and waiting up to 3 months for the loan to come through.
The Web 2.0 version uses a web page or app to capture user information. A combination of FAQs and web chat to answer customer questions, a startup that automates the Know Your Client step. Some simple rules are implemented on the lending criteria, so you either qualify immediately or computer says no. An app with electronic documents for you to sign and you're done.
While you might struggle with some of the questions on the forms or wonder what they mean by certain documents, they don't bounce you around between customer service agents or leave you hanging on the phone, so all in all these experiences are better for the customer and cheaper for the bank.
The AI Automation version uses a mix of web forms and AI chatbot / AI voice chat to answer all your questions. The Know Your Client step uses AI to automatically verify the customer. AI lending decisioning that can take in many more factors and give a better more personalised decision. AI Automation built on Kortical can catch 83% of the credit defaults that would otherwise occur using credit score. So AI Automation not only leads to a better more personalised customer experience but a more profitable business.
In the death by a thousand cuts version, they adopt a different best in class startup for each phase of the process.
The AI companies absorb your data into their models. You work with them to refine their product with your knowhow until it does exactly what you need.
They keep growing their capability and you need fewer and less skilled people to operate it. Keep repeating this over your business and eventually you have a legal entity with a brand, customer relationships, a large chain of suppliers and little else. This might sound like an exaggeration but some companies are acknowledging this and already embracing this direction.
We've all been hearing data is the new oil and in the AI world that is especially true. As companies hand over their data to leverage the AI Automation, they are actually handing over all the expertise, process and experience that's captured in that data too. While the AI companies sell you on the idea that you will get to leverage other people's data, everybody that goes along with it is effectively training their AI competitor to replace them.
In 2013 Gartner VP Peter Sondergaard said "Every company is a technology company" and that mantra is often repeated as it becomes ever more true. The "Go big or go home" play for companies that see the writing on the wall is to build their own AI Automation for their core business functions. The problem is that getting from something that works in a lab, to a revenue generating solution is hard. Gartner data shows that on average only 8% of AI projects are profitable and they take on average 2 years to get to pilot and 4 years to go live. This results in a lot of big expensive bets that don't pay off but for those who do make it, it means an unfair advantage over the competition. Data and knowhow are the lifeblood of an AI solution and help to create an unassailable moat, as new AI startups with better technology but worse data, won't be able to compete. This is why "data is the new oil" or more accurately why data is to AI monopolies what network effects are to social apps.
While building 13 AI labs should statistically get you a live AI Automation solution in 4 years, this is not a realistic option for most businesses. Hence the modern business dilemma: Death by a thousand cuts or go big.
Kortical has created a third option, you keep your data & expertise and use Kortical to create your superhuman performing AI Automation solutions across your various processes. This allows you to rapidly deploy Superhuman AI Automation and all the benefits that incurs, typically 70% - 95% of costs and build your own AI IP to create a defensible moat against the competition. It can also go live in as little as 2 weeks. In large enterprise where it takes more time to coordinate data access, IT and change management it typically takes less than 6 months. You can read use cases on some of the household names that are already using Kortical here.
How it works:
Identify use case
Typically the biggest cost centres of a business are focused around core business functions. We identify where Superhuman AI Automation can be best applied. This is typically somewhere the inputs and outputs to the worker are digital and the decision making process or insight generation is quite repetitive.
Creating the AI Automation
You then gather up the inputs and outputs, load them into Kortical and kick off the automated machine learning to iterate through thousands of potential machine learning models and solutions to find the one that performs best for your data. The performance of this step determines the business benefit, so you want a high performing solution for this stage and Kortical outperforms Google at AutoML. (Kortical inherits no rights to your data or the models created from it, this remains your IP and competitive advantage). Through Kortical's insight and your domain expertise you can further refine and improve the solution.
Validating the AI Automation
You hold back a portion of your data from the creation process, to see how the AI Automation performs on unseen data. This allows you to quantify the benefit before making any investments in integration or change management. Using the Superhuman AI Automation process, these results are typically human level or above.
Creating the AI Automation service and integration
Kortical provides a number of templates based on open source tech stacks that can be used as is or customised to deliver production, enterprise grade AI Automation. This includes REST APIs, data store integrations, workflows and tools for highly automated dev ops and ML Ops.
Human in the loop / Self learning
While AI Automation will typically take care of 70% - 95% of the workload, usually there are new inputs that are substantially different or complicated edge cases that crop up, so those cases are labelled as for review and sent to workers to be completed. The outputs from these tasks can then be fed back to Kortical, so we get ongoing feedback and new challenger solutions can be built to either deploy automatically if they outperform the incumbent champion or to send detailed reports on performance so challengers can be manually approved. In this way the AI Automation learns and improves over time.
Because Kortical includes powerful automated machine learning solution creation, it builds high performing machine learning from your data. As such it's already adapted to your business, best practice and ways of working. Typically we see human equivalent or super human performance for the cases that it handles. Data tends to differ across businesses quite massively and as such the system that is best at learning, is able to provide the best results for a business. It’s often counterintuitive to think that a widely applicable system would be able to beat a niche system but with machine learning and AI, it’s more similar to intelligence, where the best learners are the best at learning anything and that learning allows them to become highly specialised for a particular niche.
In the same way that if you want to build a website, you don't start building a web server and a database, you use off the shelf and just build the business specific part of your application. The mistake a lot of businesses are making is trying to build the AI infrastructure layer, as well as the AI solutions on top of it.
By leveraging best in class Superhuman AI Automation, you can get incredibly impactful business savings with 70% - 95% automation across key business functions in as little as 2 weeks, retain your competitive advantage and data while building a competitive IP moat and side step the dilemma entirely.
With McKinsey saying “~50% of all current work activities are technically automatable using currently demonstrated technologies.” and the ability to rapidly create this automation here, as Elon Musk said “companies have to race to build AI to remain competitive. If your competitor is racing to build AI and you don't, they will crush you,” The AI future has already started, do you get on the wave or does it roll over you?
The impact on jobs and people’s livelihood is obviously a major concern with AI Automation and we at Kortical feel very strongly about trying to ensure that AI is a net positive for humanity. Read more about our stance as well as what we're doing about it in our mission statement.
Some of the good news from what we’ve seen of companies using Kortical to drive Superhuman AI Automation is that it removes a lot of the repetitive work of answering the same questions or doing basic reviews and has the workers doing more interesting edge cases that are more engaging of their talents, so we see that worker happiness and employee retention tend to go up. As well as this we typically see that once the more routine jobs are taken care of, there is more higher value work that workers can be trained up to do. As such we’ve not yet seen large layoffs as a result of automation, companies instead seem to be preferring to reallocate that productive capacity.
Get In Touch
Whether you're just starting your AI journey or looking for support in improving your existing delivery capability, please reach out.