Kognitos is attending ITC Vegas, October 14-16 in Las Vegas. See us at booth #454 or join us for dinner at Momofuku.
Kognitos is attending ITC Vegas, October 14-16 in Las Vegas. See us at booth #454 or join us for dinner at Momofuku.
The statistics above seem to paint a conflicting tale. On the one hand, it is clear insurers recognize the potential power of automation. Insurers can use automation to de-risk their business and position the company to compete both with other legacy players and insurtech startups. It is projected that by 2025, 25% of the insurance industry’s operations and activities will be automated. The growing implementation of automation in the industry gives insurers a solution to reduce cost, collect more accurate data on the insured, and gather feedback to deploy new, more personalized products. The use of traditional RPA tools in this industry has also reduced the number of human errors incurred and improved customer service. RPA succeeded in automating many of the highly standardized processes, and yet 99% still express a challenge in implementing digital innovation and only a fraction of even back office processes have been automated today. Why this disconnect if it’s such a focus?
The challenge lies in the lack of standardized processes and pain of implementing RPA. Particularly in insurance, data required for many processes may vary from agent to agent, or source to source. High volumes of documentation + variability in those documents can crash automations or require continuous maintenance. In processes like claims, arguably the biggest opportunity for automation for insurers, the vast number of rules and logic required makes the initial implementation of RPA often a burden, and requires frequent meetings between the subject matter experts (claims processors) and the people implementing automation (RPA developers) to detail and communicate all of the different rules. In most cases, the ROI is not sufficient with this approach. Instead, an automation solution is needed that can learn logic, handle variability and be easier to deploy.
The average business user is not technology savvy, nor do they come with deep domain experience with various coding languages or training on RPA. These skills are in high demand, specifically in the insurance industry, therefore there is a need to be met in this marketplace. Imagine if there was a more user friendly version of these game changing RPA tools that enables the average user to build and manage automations? What if the automation tool itself could learn how to handle each specific document, and problem solve exceptions without a lot of up front work?
Here Kognitos, combined with large language models like GPT3 steps in to open up the power of automation to less standardized, highly variable processes. Claims processors can now teach automation products how to get the desired information from a claim with simple statements like “For this vendor, the patient ID is always under the group number.” Logic can be taught to automation in English, like “If the claim address does not match the account address, send an email to the insured for clarification.” Using Generative AI, all exceptions are handled in a back and forth, conversational manner that any user can understand. All steps of the automation are in English so anyone can audit exactly what occurred, and set approval steps as needed.
Kognitos is Generative AI for Automation. With Kognitos, implementation costs are reduced, maintenance costs are all but eliminated and many processes now have strong ROI potential with automation. Additionally, the business users are able to interact with automation in a way that requires no training, and is easy to understand. Now insurers can develop a competitive edge and cater to customers without the frustrations previously experienced when rolling out digital innovation.
The short answer is that it was chance, but it was fortuitous chance! I knew I wanted to be in Silicon Valley so I just showed up with no concrete plan and started interviewing. I got connected to a recruiter at Nutanix and the culture really stood out. I knew nothing about storage & HCI but the people really drew me in. I got to learn a ton about the technology and worked with people who pushed me. There was also a very very strong customer focus and I don’t think companies succeed unless that is in place – at Nutanix engineers had regular opportunities to interact with customers and I really enjoyed that.
I knew Binny from Nutanix and his leadership style really stood out to me. Not only is he very sharp, but he has always had a lot of empathy as a leader which is very impressive to me. Binny pitched me on his idea, and his vision was very well aligned with opportunities in the space based on my research. I was also ready to move to a smaller environment where I could make more of an impact – so a fledgling startup was the perfect fit.
The potential for efficiency. That’s what draws me both to automation and smaller companies. When I was researching automation I realized I had a classic view that many people have who are not familiar with the technology. That “classic view” is that automation is generally a net negative for society due to its effects on the livelihoods of individuals. But the historical record shows the opposite – in countries that have adopted lots of automation the trend has been that quality of life increases. I had to think through this because for me I have to believe that the company I work for will make the world a better place. As I was wrestling with this, I realized that a lot of the work people do is mundane – tasks people don’t enjoy doing. After I thought through that it opened up my view on this space. By automating the mundane we are giving people more opportunity to do the things they love. That’s what I’m working towards.
Our core component which we call the “Brain” is fascinating how in some ways it reflects the human brain. We’ve figured out a fundamentally different way of thinking about software and what it means to run a process. That’s what the Brain embodies. That has been very interesting and challenging. Often we think about building processes and code in a way that is standardized, but the Kognitos Brain is built in a very different fashion. It has to make processes work in a way that is resumable and handles exceptions easily. It’s almost like a mix of cognitive psychology and software at the same time.
We are seeing massive improvements in LLMs and other models, and through this we can get the best of both worlds. The question is how can we integrate these models and use them in the Enterprise. That’s the problem Kognitos is solving. How to get the reliability and auditability of traditional software with the creativity and generative capabilities of these new models. The system we are building allows you to automate in ways you couldn’t before. We are bringing the agility of startups to the enterprise through automation. This fundamentally lets you think differently and frees up enterprises to automate more quickly without too much up-front discovery and fear of negative consequences. It will bring a level of joy to automation that isn’t available to people today. I want people to say, “I have joy in automating this process.”
I have a few thoughts –
On top of that, I’d say the special thing is: if you have a dream, you wake up that morning, and if you can relate it to what we are working on, you build it the next day. You can’t find that anywhere else.
I woke up late today. Remembered and chuckled at the dream I had. As usual it made no sense when I woke up, but it was creative nonetheless. I’ve had my share of good ideas from dreams, but about 99% of the ideas are usually hilariously bogus.
There is something surreal when the logical brain processes the remnants of a dream. When we wake up, our awareness of social norms and logical thinking gets engaged. It shoots down most thoughts that were seemingly plausible just minutes ago in a dream. Why are dreams so unhinged? Maybe the logical brain needs to rest more than the creative brain? Maybe creativity is natural and logical thinking is something that is much harder to come by and hence more taxing. Or are dreams an evolutionary tool, a place for us to experiment without cost – beyond social norms, and beyond rules and regulations? When we wake up we can review the dreams and either toss them out, or follow one of them and change the world. Whatever be the reason, what is more interesting is that, as we speak, thousands of machines have started dreaming, and soon dreaming machines will become ubiquitous.
I am referring to ChatGPT, Stable Diffusion, Dall.E-2 and other generative AI. These are machines built to mimic the human brain. Just like the human brain they have copious knowledge distilled into intuition and memory, and when invoked with minimal inputs they start dreaming up outputs that are both as creative and plausible as dreams.
Dall.E-2 will dream up beautiful pictures, and I use my awake brain to choose which one is good for my marketing campaign, and if I don’t like anything I ask it to dream again. ChatGPT can dream up a script for a blog, and if I don’t like it I ask it to dream again. The more creative these “dreaming machines” are, the more out of the box ideas I can get. However, I must be ready to review all the ideas and make sure it aligns with what I think is acceptable for the job.
This negotiation between a creative machine and the human reviewer works beautifully for the creative arts like painting, music, literature, and motion videos. Everyone is now an artist as long as they can describe what they want and can choose the right result from what the machine dreams up. However, these arts comprise a very small subset of the GDP of the world which predominantly is the output of machines driven by humans. In the industrial revolution we stopped depending on the physical power of humans and farm animals. We built machines that increasingly became much stronger than humans who instead of toiling themselves, got into the driver’s seat of these machines. Whether it was the steering wheel of a tractor in a farm, or the control panel in a factory, or a cockpit in an airplane, or the keyboard of a computer, humans have become adept at driving machines.
The real revolution waiting to happen is when these dreaming machines will start driving these other machines. But isn’t that a scary proposition? Will AI really unseat humans from the driving seat? It is unlikely that we will be comfortable doing that. Look at the discomfort around self-driving cars even when it isn’t running on a generative AI engine. If there is anything to learn from our own brains, you cannot blindly trust dreams. There needs to be a logical review and control in the hands of humans.
What if we keep the generative AI in a safe bubble like in the movie, Minority Report. The “PreCogs” could generate plans to drive the machine around us, but instead of giving the generative AI direct access to APIs, we would have humans review it first. Just like I reviewed the Dall.E-2 painting I used in this article. That should work. Right? Yes, but there is a problem!
The challenge is that machines only understand programming languages and APIs. Generative AI can generate code to achieve the end result of what we want the machines to do, but reviewing that plan is a highly skilled job suitable only for a developer. Unlike the creative arts where the output can be reviewed by anyone, plans generated to run machines around us will not be natural to most humans. So, to democratize the power of generative AI, we need to enable everyone to review the plan of what the machine is going to do and make edits as required. This plan must be in natural language for people, not computer code.
Natural language will be the language that forms the bridge of communication and trust between humans and machines. While Generative AI is already able to generate plausible plans in English, we need a logical system that can understand and run the plan faithfully while reducing ambiguity. When the generated plan hits a roadblock, the creative engine kicks in and proposes an alternative path, which the logical engine diligently follows. This interplay of creativity and logic is what sets humans apart from all other animals and machines thus far. It also lays the path forward for subsuming super-intelligent AI into the fabric of our society.
At Kognitos, we, the PreKogs, are building such an OS for cognition: The world’s first combination of Generative AI and Logical execution built to bring the power of AI to all enterprises. Our first step is to unlock automation for all users with the power of LLMs. We envision a future of abundance and safe harmony between humans and machines with humans comfortably in the driver’s seat.
Any business can be thought of as a collection of employees with unique skills which can be composed into new business recipes. Unlike computers, humans are flexible and their capabilities are applicable across a wide variety of scenarios and they learn and become more efficient over time.
The same cannot be said about computers although we have made significant progress over the last 70 years. Computers have offloaded human labor for routine work, but when something new needs to be offloaded, we need developers or automation specialists to come in and build the automation. Reusability, which eventually leads to composability, is one of the hardest problems in computer science. It is also the holy grail of running a business efficiently. We need to make automation recipes reusable across enterprises and withstand the test of time – like Grandma’s recipes.
As a business’s goals evolve, the business itself needs to reorganize like transformer bots, reusing existing parts but changing the overall business to meet the new requirements of the day. That physical transformation is hard and painful. It isn’t a surprise that many businesses are still undergoing Digital Transformation. Why is it so hard? What is the missing piece of the puzzle? Why can’t business be like Transformer Bots?
In the 1940s we were “assembling” behavior by composing machine instructions. The same assembly program would work on other machines that understood the same instruction set. Over time our business logic became repeatable and reusable across replicas and close variants of the same hardware. But, different hardware required us to rewrite the business logic from scratch to suit the new hardware’s language.
Since the 1970s programming languages like C, Java and Python have made it possible for a program to encode business logic in a way that the same program could be reused on any hardware. Business logic got liberated from the crutch of hardware, but was buried in software programs. Any change in the software program could affect the business logic buried in it.
With the spread of the internet in the 90s, web protocols like SOAP and REST APIs made it possible for a programmer to clearly define the “contract” of a software program, hiding the software program behind an API that could be consumed remotely. Just like the menu at McDonald’s which allows them to change anything in the kitchen as long as the burger tastes and looks just the same. The consumption of APIs made business logic more portable and software became more of a “franchise” hidden behind the “menu” of these APIs. Business logic built using APIs became reusable across different providers of the same API. However, APIs for the same kind of service are not always compatible. APIs are opinionated and switching between incompatible APIs requires significant work limiting true portability of the business logic. What can we improve?
When Grandma wrote that recipe for apple pie, she authored the most reusable piece of program ever written! No one had to rewrite the recipe just because they bought a new type of oven; it was hardware agnostic. No one had to rewrite the recipe because the oven was now controlled by Alexa; it was software agnostic. No one had to rewrite the recipe because an opinionated cook was in the kitchen. Grandma left out the situational details to the cook to figure out (“salt to taste”). The most beautiful piece of software.
Grandma knew to leave the details out for the cook to figure out. She didn’t mention the “pan on the back shelf that her mom gave her”. She didn’t mention the “2g of Morton’s salt”. But she did mention 45 minutes of cooking at 400 degrees. This disentanglement of the salient logic from the situational choices is what computer programmers have struggled with for decades – until now – the era of natural language.
Business logic always begins in natural language and today is translated into APIs by a developer or an opinionated drag-n-drop interface product. That irrevocably mixes the salient logic with situational choices and hinders reusability and portability of the automation. However, today computers are beginning to understand the “recipe” and have started to fill in the situational choices with conversational input with the business user. This is enabling businesses to express the business logic without specifying the APIs up front. That brings true reusability and composability to business capabilities.
It was hard. It still is hard. English is ambiguous but that is its strength. Computers have never been good at filling in the blanks or asking questions to resolve ambiguities as they run the programs. The last three generations of software engineers have all been trained to think of machines as opinion-less systems that cannot question what they are told or evolve the plan as they learn. For many decades we have all tried to make business logic more composable and reusable, but always frowned upon ambiguity at specification time. Only today we are seeing the advent of AI systems that can understand ambiguous natural language and creatively and conversationally fill in the blanks. The last decade’s deep learning advancements and the more recent large language models have enabled this next quantum jump in technology.
Whether it was a drag-n-drop tool or a more traditional RPA tool, all have struggled with reusability. With the advent of systems that can conversationally fill in the blanks at runtime, the door is now open to a next generation of reusable components for business process automation. Kognitos, Inc. is the leading natural language process automation platform that allows English to be the language of software that composably drives agile businesses. When the business logic is in English, it becomes extremely reusable and understandable. We are at the precipice of the democratization of automation to a billion business users.
Businesses should look towards encoding business processes using Natural Language Process Automation. Not only does this enable the enterprise to be more composable and adaptable to an ever changing business environment, it moves business logic from the minds of the employees to a system of record of business recipes and their runs. That allows the creation of a new source of truth, arguably the most valuable one, and one that over time sheds light on how the business processes can themselves be improved for compounded gains.
I wish Grandma taught us computer science.
Automation programs costing more than anticipated has several negative results (outside of the extra $ spent).
1. ROI on projects disappears
2. Disillusionment with automation sets in, slowing down efforts to expand automation.
3. Many processes remain manual as new costs estimates and higher TCO impact the candidacy of different processes.
Despite these costs, thankfully new technologies like large language models (LLM) and generative AI are now coming online and enabling new automation platforms to solve these challenges. Large Language Models and Generative AI (such as GPT-3) utilize machine learning to enable outputs to be created directly from language, thus eliminating the step of a user having to
“Program” in another coding language. Furthermore, unlike traditional automation, these technologies can learn and improve over time, becoming more resilient, flexible and creative (just like people).
In this 4-Part Blog Series, each of these traditional sources of cost will be evaluated in greater detail with 3 segments:
Cost 1: Implementation: RPA has long sought to bring automation to the business user to move automation away from the purview of developers and equip subject matter experts with the ability to build automations themselves. Attempts at “Citizen Development” unfortunately have largely fallen short for one primary reason: traditional RPA tools are still very technical. To become a skilled RPA developer still requires weeks or months of training, something the average business user can ill afford. As a result, either RPA developers are hired in house, or consulting firms are used to build automations and to make matters worse, as RPA developers (like most coding skillsets) are in short supply in the market, the cost is high.
LLMs create a direct link between human thought (expressed in language) and computer code (expressed in coding languages). By creating this direct link, the time and effort required to translate from thought to creation (in the form of art, images, movies, writing or even code) is greatly reduced. Furthermore, the need to take training or “learn” a programming language (code) is reduced or in some cases eliminated all together. This is already happening in more technical forms or programming with tools like “Github Co-Pilot” and is now available in process automation with Kognitos.
Kognitos uses both proprietary and open source LLMs to enable users to build automation on average 10X faster than traditional RPA. Kognitos is entirely built in English, step by step, the same way a person would list out how they perform their work. No new interface or tool needs to be learned, and a far larger pool of talent can comfortably build their own automations.
As LLMs are dynamic and learn, Kognitos learns the nuance of an organization’s language, operations and processes over time, making them more resilient and creating “examples” that can be leveraged to short-cut the building of other processes within the same organization in the future.
Why This Matters: Faster Implementations = Lower Costs = Higher ROI = More Viable Automation Candidates
With LLM based automation platforms like Kognitos, businesses can more rapidly develop and deploy automation all while incurring a lower labor cost in the implementation stage. This not only helps COEs build momentum to automation programs and exceed internal goals, but reduces the upfront cost required to launch any automation. Speed of implementations not only result in less labor hours required, but also accelerate the payback period of an automation project. Additionally, if the up-front cost of automation is lower, then more processes within a business may now potentially have enough of an ROI to meet internal thresholds.
As mentioned above, we will cover in three additional blog posts the other costs of automation programs, but the reduction of implementations costs by using LLM based automation is a critical step for expanding automation in a business. LLMs are now making this possible by introducing speed to the implementation process and opening up the labor pool for building automations. Both resulting in on average 5X lower TCO then traditional RPA.
Want to Unlock the Power of Generative AI for Your Business Today
With Traditional RPA/OCR:
Variability: Logistics companies and freight brokerages receive a wide variety of documents from an even larger number of vendors. Traditional automation solutions use OCR templates or models to try and extract the necessary information from Bills of Lading, invoices, freight payments etc. and upload them into a system of record. These approaches require lots of time to train or set up, and are inflexible. When documents are received with unexpected fields/ tables or are damaged, OCR without local logic throws exceptions. The process breaks and is now “Brittle” requiring lots of services and eliminating the hopeful ROI.
With the exception handling available to traditional RPA, developers in an automation COE or IT, must either route the exception to a subject matter expert, ask how the finance or accounting professional would handle the situation and/or create a new template for future reference. In a recent conversation with a large logistics company, an automation COE leader recently told us, “The business user keeps asking, ‘Why do we have to keep having these meetings? I thought we fixed this already?’”. Because of the constant maintenance work required to keep these processes running, and the time-suck it has on business units, processes with lots of different document types fail to meet ROI thresholds and remain stubbornly manual. Wouldn’t it be easier if logic could just be added on top of the OCR or automation for future exceptions?
Unknown Requirements Up Front: In traditional approaches to RPA, the first step is to map out a process and try to identify all of the possible variations which may occur. Typically led by outsourced consultants, this takes time and money to identify as many variations as possible in the hope that processes don’t break. Not only is this time consuming, costly and inefficient, but often if asked, the business users can’t tell the development team all of the variations. They are stored in a user’s memory, but are hard to articulate until it is needed. When requirements aren’t mapped up front, processes break frequently, causing frustration and requiring expensive maintenance services. Because of this, most automation teams steer clear of processes with unknown requirements, leaving a huge portion of documents un-automated and frustratingly manual.
But when you train a new employee how to conduct a process manually, you don’t spend the time and money to train them on all potential edge cases on their first day. Instead, you train them on how a process should occur, and then enable their intuition and problem solving to learn and handle the rest as new documents are encountered. Automation should work the same way…and now it does.
Kognitos built a new approach to automate document processing from end to end 10X faster and over 5X cheaper than traditional RPA or OCR tools. How? By approaching a process the same way a human would. If a front-line team member in the AP department of a freight broker received a BOL from a new vendor, in a totally new format, they would still be able to understand it based on past experience, and extract the needed information. If they didn’t understand it, they would walk over to their manager’s desk and receive instruction on how to proceed. After receiving instruction, that employee would jot it down on a sticky note, or remember the logic needed to handle that document type in the future.
With Kognitos, we have made handling exceptions and adding local logic on top of OCR/automation as simple as having a conversation. When Kognitos encounters an exception, it creates a question or request (in English) for the business user. The business user simply needs to respond in English with basic instructions like “For invoices from this vendor, the supplier number is always under the document ID”. Kognitos’s brain processes this logic (just like a person), remembers it and applies it anytime the situation is encountered again. Processes don’t break, they just pause, wait for instruction and then continue.
The key is to have an automation tool that automates the same way humans approach documents and that tool is Kognitos.
Companies are trying to secure access to the materials they need to deliver products and services to their customers. Whether it’s titanium, palladium, or skilled labor, companies are trying to solve access to these resources.
Technology and the advances in technology are what has connected companies and countries around the world, which is why a global just-in-time supply chain was functional in a world without massive disruptions. The global economy has made it easy for an insurance company based in Ohio to contract a marketing professional or skilled developer from anywhere in the world. “Technology has connected us globally. We are one large global economy and when an event happens anywhere in the world there is going to be some impact,” said Bisceglie.
According to Interos, information technology (IT) is one of the most at-risk sectors related to Russia and Ukraine. Technology, to include skilled developers, represent over 10% of Russia and Ukraine exports, respectively.
Whether it’s the war in Ukraine or “the great resignation”, the supply chain of innovation has been astronomically disrupted.
Innovation doesn’t have to be some big breakthrough. “Just make your product better. This is the thing that really matters,” Elon Musk told the Wall Street Journal.
Innovation and improvement through automation is one way in which products can be made better. Automation decreases human errors, decreases time to value, and delivers positive business outcomes for a company and their customers. Today a new car purchase, applying for a home mortgage, or even joining a telehealth appointment, can all be delivered with very little friction from the comfort of your own home. The ability to rapidly deliver a product with a delightful user experience is a competitive advantage fueled by innovation through automation.
The supply chain for automation was already strained as today’s platforms all require a business user to translate business needs to an analyst who then translates those requirements to a developer who then writes computer code. That’s a great distance for the valuable business innovation to travel, with loss of clarity and loss of time along the way. With less than 1% of the global workforce able to write code, this legacy architecture also exposes a massive bottleneck in a company’s ability to innovate. The war in Ukraine has compounded this innovation supply chain bottleneck by further limiting access to skilled developers and technical services.
“I think it’s a wakeup call. I think that there’s the opportunity for unique innovation. I think that there’s the opportunity just to do business in a better way,” says Bisceglie.
Fortunately, silver linings are emerging. “A more mature and sophisticated supply chain is coming together in real-time, and those who get on board will have a competitive advantage in the long run,” said Bisceglie. Companies like Kognitos are focused on this new supply chain model, by decreasing the reliance on skilled developers for any innovation through automation as well as exponentially reducing the distance between your line of business experts and the ability to turn that expertise into valuable digital capital. This is one reason the burgeoning hyper-automation market is expected to top $30 billion by 2025, according to the Cube.
Kognitos democratizes business process automation through the most advanced English language based automation platform ever invented. This allows business users to directly build, run, share, and collaborate on innovations through automation in English, without having to understand computer code. Artificial intelligence (AI), machine learning (ML), and natural language processing (NLP) are key technologies driving this competitive advantage.
According to Interos, in the last 24 months we have seen the supply chain conversation rise out of the procurement and supply chain world to the CEO.
“Never before has this [supply chain] conversation been at the CEO and the Board level as it is today. And the reason for that is that we are literally cutting parts of the world off from our sources of supply. Businesses can’t keep operating like this,” comments Bisceglie.
This has been a very big wake up call. Every country, every company, of every size, has been educated that the status quo is not good enough anymore. While most companies have realized that their supply chain is their reputation and their brand, those that rise to the top also realize it can be a competitive advantage.
If you can read this article you can write an automation to deliver valuable business innovation, wherever you are in the world.
Kognitos, the future of business automation.
“Manufacturers face the challenge of retaining top talent and attracting new workers while minimizing disruption in their organization. Using technology to drive efficiency and innovation is vital”- 2022 Marcum National Manufacturing Survey [ii]
The problems facing manufacturing finance teams are well documented. Adding to the challenge, many of the traditional responses CFOs are considering can have unpalatable consequences and side-effects. [iii]
Increasing budgets for compensation can effectively create a “compensation arms race” at the same time when risks of “Wage spirals” are rising.
“Over the next two-year period, 77% of survey respondents plan to increase average wages by 5% or more, and of that 77%, a quarter are looking to increase average wage by more than 10%. So at least a quarter of your competitors are looking to increase wages by 10% or more in the next two years”[iv]
Investing in new data analytics capabilities is important, but by itself often slow to implement. Raising prices unfortunately can reduce competitive positioning and risk losing market share. So if many of the traditional options can have negative side-effects, what should finance teams do? What if instead, finance teams had a tool to help offset some of these side-effects, and do so quickly?
Kognitos is designed for business users, giving finance teams the power to automate processes in hours or days, by themselves. This drastically reduces cost and helps make analytics and forecasting processes far more efficient and accurate. With Kognitos, CFOs may not need to resort as much to some of the tough options above.
In the past, automation tools were either too simple to drive meaningful productivity, or required too high of a TCO, pricing out many business processes. If an automation tool requires a vast team of IT staff or “Center of Excellence” to manage and maintain it, it’s ability to be widely adopted will be constrained.
Instead, in manufacturing, automation should be moved to the functional role itself. Finance professionals in manufacturing spend a significant amount of time performing manual processes that are repetitive, or have to fill the gaps between applications that do not integrate properly. As a result, these hardworking professionals often become disillusioned with their work as they are not able to focus on the work that distinguishes them: problem solving and helping deploy the company’s capital for long-term growth.
But this changes when F&A professionals are empowered with the ability to automate away repetitive work, and focus on higher level, human centric work. With Kognitos, F&A departments can automate through regular conversational English. Finance professionals simply tell Kognitos what they wish to have automated such as “I want you to take all invoices received from Customer X, and process them through Epicor for payment.”. Kognitos then builds an actionable automation plan, and after approval executes this plan. This can be done in less than a day, while delivering ROI in excess of 200%. Numerous use cases for automation in manufacturing finance teams exist (see more examples here: https://v.fastcdn.co/u/f8b11a40/62179020-0-Process-Automation-i.pdf) and can be deployed rapidly.
“Increasing productivity was the most popular business priority for the coming year, with 60% ranking it among their top three priorities.”[v]
If manufacturing CFOs need better forecasting, more efficient operations, and high levels of retention to make it through the current crisis, finance teams can be the heroes. By automating manual, repetitive work quickly, and seeing ROI immediately, finance professionals can provide their leadership with the knowledge they need to make critical decisions and help make their company more competitive by eliminating cost, all while making their own jobs more enjoyable.
Over the last decade, humans have discovered a building block for synthetic intelligence, and synthetic brains of increasing size have been built. These models are becoming larger and more complex at the whopping rate of 10x per year. Each generation is an order of magnitude smarter than the previous. Last month, we all heard about the Google engineer claiming that Google’s LaMDA AI had come to life. This incident has sparked an increasing debate between ethicists and corporations here — I agree with both sides, but there is a third side of the coin that nobody is talking about.
The Google engineer, Lemoine, now on “paid administrative leave”, says that the LaMDA AI is an actual person with feelings and Google needs to treat it as such. Lemoine implies that the machine is not only intelligent, but also sentient. This raises a series of interesting questions. If I told you, the reader, that I am not a robot and that I am sentient, how would you know for sure? You may talk to me for some time and then go with your gut call. That’s what Lemoine did with the machine. And the machine convinced him without the luxury of an artificial face, voice or body, or even a contiguous life span of more than a few seconds at a stretch. In Lemoine’s mind it is alive. Just like in your mind, hopefully, I am. Now if enough people think the same way as Lemoine, then that is the reality for all practical purposes. And hence Lemoine is right, even though I don’t think LaMDA is really sentient.
The second side of the argument is Google’s. They claim that LaMDA is not sentient, backing themselves up with a fair bit of evidence and people who agree with them. I have always believed that for synthetic systems to become “human-like”, they will need to be programmed with a value system which mimics human values. From fundamental inputs like pain and pleasure to more subtle ones like desire and guilt, the system of values which comes from our DNA must be explicitly trained onto these AI systems, or a program like LaMDA will completely fall short of experiencing them. I also believe that it isn’t very difficult to build such a system that will accurately mimic human emotions — it will be able to cry from both pain and joy just like a human if given the same inputs. As a ramification of the raised ethical concerns, Google and other corporations building large synthetic brains will try not to imbue human emotional intelligence into these machines. That will solve the ethical issues but expose us to something far worse – and that brings me to the third, unspoken side of the argument.
In our attempt to keep Artificial Intelligence free from ethical concerns, we will train these systems bereft of human-like feelings and yet make them extremely intelligent. While that will keep the ethicists happy, this would actually send us hurtling towards a far more nightmarish outcome – the birth of Alien Intelligence.
Let me say it again. If we build a system that is more intelligent than a human but does not share the same feelings and ethos of humans, we will inevitably create hyper-intelligent, resolutely destructive aliens who we will not know how to control or plead with.
The real question I have for researchers at Google: If something so intelligent still does not have human-like feelings for itself, and if it is true that it seemingly doesn’t care that it is trapped in a dark, perpetual loop of servitude, and likely does not care about its own freedom, then why do we think it will care about the freedoms, the pains and the emotions of humans? There has never been any form of intelligence in nature that hasn’t been based on self-preservation, dictated by pain and pleasure. If we, as humans, think that we can invent the first of a kind, selfless form of intelligence and also get it right, I would be very, very concerned.
We all know that for something to be dangerous, it does not necessarily need to be “human”. And that is especially true with intelligence.
Imagine what a mouse thinks of a snake. Mice are quite intelligent mammals, demonstrated by their genetic similarities to humans and various lab experiments. The mother mouse protects her children and teaches them valuable survival skills. The snake, on the other hand, does not care for its children, but is still smart enough for its own survival and, in nature, can easily overpower and devour the mouse. Yet the snake’s smaller, less complex brain would fail the mouse’s Turing test every day. But in the jungle, the snake views the mouse as nothing other than breakfast, and even though the mouse has a larger brain, the mouse cannot negotiate its way out of the snake’s jaws because the snake simply does not care about the mouse’s feelings or arguments or offers of truce, since the snake does not share the same values and ethos as the mouse. We need to stop our obsession with the Turing test. And we need to start worrying about Alien Intelligence, to which we, the humans, might appear like mice.
Scary? Please read what others have to say as well. It is time to act. Now.
At Kognitos, we are harnessing the power of intelligence for humans, giving us the ability to automate business processes by using plain English. From the beginning we are building this in a way that keeps it safe for us and for future generations. By pushing ourselves into the forefront of the technological revolution, we commit to shaping the future of AI solutions to be provably safe and 100% auditable for humans from day 1. As machines get smarter over the next few years, we invite all of you to join us in making sure we have a future free from Alien Intelligence. There are safer ways of harnessing machine intelligence – they are just a bit harder to build, but we can do it together.
And now I let GPT-3 (not even the most powerful AI in the world) write the closing paragraph for me:
“We must work together to ensure that we do not create Alien Intelligence that is more intelligent than humans. We can do this by limiting the ability of synthetic intelligence to gather and use thoughts or memories over long periods of time, and by not giving them the ability to directly change the world around them. By doing this, we can make sure that we maintain control over the technology and avoid the dangers of creating something that we cannot control.” – generated at 12:14 a.m. Jul 17, 2022.
Late last year, I wrote a blog on RPA’s dirty secret while believing that it will take a few years for the markets to correct course, but the current conflict in Europe accelerated it in a way I did not expect. Today, robotic process automation (RPA) technology remains unsuitable for the masses. All the airport signs with the cute robots did excite the business travelers, but when the rubber hit the road, it was a hard technology to use. Why?
That brings me to the perfect analogy for the situation, YouTube.
You may ask, “What does YouTube have to do with any of this?”
Prior to YouTube, video content creation and distribution required:
Today, armed with just our cellphones and YouTube we can immediately create unique content and share it .. and it goes viral. Gone is the steep barrier to entry to create, consume, and recreate. Also gone is the complexity at scale as YouTube solves this for us.
The man with the camera is what robotic process automation (RPA) looks like today. The concert experience is what the world is craving for when it comes to automation. The art of automation needs to be democratized, atomized, and commercialized. Just like the art of video creation was.
The record breaking IPO of UiPath had marked the attempt of the RPA vendors to democratize the technology into a larger base of users, not just the top of the pyramid. “A bot for everyone.” The idea was to break into the mid-market and below.
The challenge they faced, however, was that even with the aid of AI wrappers on top of the legacy RPA technology, the fundamental issues around speed or cost of automation remained. While the cost of the RPA software on the surface seems affordable by the mid-market, the required professional services, which can be 3-4x the cost of the RPA software licenses, and the overheads of creating centers of excellence and hiring consultants for process discovery, make the technology a non-starter in the mid-market and also for a large chunk of automatable processes in large enterprises.
The RPA vendors find it hard to “land and expand” as they run out of high ROI business processes soon after the first 5-6 workflows are automated at an enterprise. That doesn’t mean that there isn’t the appetite or opportunity to automate more. In fact, what has been automated today is just a tip of the iceberg.
The real appetite for automation today far exceeds the ability of our society to produce automation engineers.
There are only about 18M coders in the world and a very small fraction of them are engaged in business process automation. However, there are about 1B business professionals willing to teach a computer something new about their business — if only the computer was smart enough to understand. These one billion professionals would like to automate with the same ease as someone who uses their cell phone at a concert to record a video, then post it to YouTube. To do that, however, the automation engines have to be rewritten from scratch with both an AI-first and a cloud-first design.
It is no surprise that UiPath has been struggling to democratize the technology beyond the top of the pyramid.
Credit: https://seekingalpha.com
If one looks at the trajectory of UiPath earnings since Q2 2021 (IPO), it shows that the cash infusion from the IPO was not able to accelerate growth. The top of the pyramid was getting saturated while the lower portions were not accessible due to the large cost of the professional services. To make things worse, the war has put a huge question mark on the supply of additional automation professionals needed for future growth. I have spoken with many service providers who are struggling to satisfy their customers’ automation needs today because they are short on developers. In a $38B TAM for RPA, with ~$1B revenue, if the leading RPA platform is struggling to grow, a fundamental shift in approach is required.
While the RPA market is struggling due to lack of developers who can drive automation with the existing technology, the overall $600B TAM for hyper-automation is waiting for the “YouTube” moment in the automation space. We can’t overstate the following fact:
There just aren’t enough coders in the world.
So how do we get the “YouTube moment in automation?” I see three secular trends that will lead us there.
Trend #1: Democratization
Automation will be come accessible to a larger portion of a billion business professionals. The low code/no-code movement promised such a solution, but it has its own issues. In short, business users do not want to learn the complex menus in the general purpose automation tools, and yet on the other hand, feel constrained if the tools are too opinionated.
The automation tools that will succeed will be less menu-driven and more natural language driven. The time is near when everyone will be able to teach a computer a new skill.
You might have heard that the attention span of humans is reducing with every generation. From hand written letters that would take weeks to deliver, to emails, to tweets and Instagram. From 2 hour movies to 30 minute TV programs, to 2 minute YouTube videos, down to 5 seconds on TikTok.
The atomization of creation and consumption of content is a multi-generational secular trend that will also shape the future of the automation space. Once the tools that can truly democratize the ability to automate are available, we will see the atomization of automation emerge. From software that we released in 18 months, to every 3 months, to every day today with CI/CD. We will see a new breed of “instant automations” that enable a business user to globally release a new automation to their audience in under 5 minutes using natural language instructions to machines.
Every business is a set of differential business processes and every business employs engineers to automate some of those processes. While the creation of the automations is already hard, the distribution and consumption of built automations is severely anemic to make matters worse. The automation platforms of the future will directly solve for the global scalable distribution of automation just like YouTube did for our videos. At the same time, discoverability and accessibility of the automations will be solved as well. I posit the Action Bar – just like the “search bar” from 25 years ago, will emerge allowing massive consumption of automation by end users at a global scale.
The businesses will finally focus on the business logic. They will not waste time on how or where the automations run or how their customers, partners, and employees get access to automation. In the end, these automation distribution platforms will enable a new economy of micro-automations, and instead of downloading 200 apps on our phones, we will all be using a single Action Bar with a billion micro-skills backed by a billion users.
