Discover more from Investment Ideas by Antonio
UIPath: Palantir Has Competition
Edited by Brian Birnbaum.
And here´s a link to my original Palantir deep dive: “Palantir: The Most Important Company in the West.”
1.0 An Unlikely Contender
2.0 Key Business Dynamics
3.0 The Innovation Stack or, Dealing with Microsoft
4.0 Financials 5.0 Conclusion
A brief summary of the deep dive, with the full write up below.
When analyzing tech companies, many get lost in technical details. However, companies compete in the market place according to the job they do for customers. Although at first glance UIPath seems like an inoffensive click-automating company, it is actually a serious Palantir competitor.
UIPath´s ability to develop ontologies is no where near that of Palantir´s, but it does have more liquid distribution capabilities that together with a high retention rate, position it to iterate its way towards more valuable forms of automation. I believe that wherever UIPath deploys first, Palantir will have a tougher time than otherwise acquiring that customer.
Qualitatively, management seems world class to me and the company has an air of excellence to it, with a proven capability to constantly improve its product. UIPath has, however, rented out a place in the Microsoft ecosystem which endows the thesis with a great existential risk. However, there are reasons to believe that UIPath´s consistent focus over 15 years have yielded some form of a process moat.
UIPath is at the midst of a financial inflection point, with cash from operations now positive as a result of the platform´s continued horizontality over time: customers start by deploying the solution in one division and then extend it to others over time. As such, UIPath exhibits clear operating leverage which going forward can be meaningfully compounded by the development of two technologies:
Semantic automation: it will enable UIPath to understand unstructured data in documents and thus, automate workflows of much higher value.
Conversion of natural language to code: it will enable the further democratization of the platform.
The two levers above should over time translate into higher automation volume/value at a marginal cost and thus, would hypothetically translate into higher operating leverage.
Further, UIPath has a robust balance sheet with $1.5B+ in cash and equivalents and no debt. As it continues to gain operating leverage, if its cashflow profile continues to get healthier, it could meaningfully derisk the thesis. At a price to sales ratio of 8, the market is not entirely ignorant of the potential of this company.
1.0 An Unlikely Contender
UIPath is a Palantir competitor, in a rather non-obvious way.
Palantir´s ultimate function is to help organizations make sense of their data by automating insight generation, which optimizes operations. UIPath simply watches computer screens (via computer vision, primarily) in a given organization, observing what tasks users perform repetitively and thus eventually automating them (known as RPA, robotic process automation), effectively emulating human users. However, if we remove ourselves from technical considerations for a second, they both actually do the same job for customers.
UIPath has started out in the seemingly narrow domain of RPA, but fundamentally they just pick up data and generates insight, to then automate processes, much like Palantir. Although Palantir´s ability to create and operate digital twins is far ahead and perhaps not even comparable, UIPath has a clear advantage in the distribution side of the equation, which over time may give it priority access to data: the key raw material of the value chain.
Many organizations will likely prefer to begin their automation journeys in a more nimble and lightweight fashion, even if they pay a hefty price for not going all in right at the start : such is human nature. Further, the way Palantir becomes a much more valuable company is by productizing its offerings fully and thus decreasing the cost of deployment / time to value (which it actively works on) so that it can move down market as well as it does up market. But, I suspect acquiring a client will be harder for Palantir if UIPath gets there first.
The details of the situation are as follows:
UIPath is much cheaper (and seemingly faster) than Palantir to deploy. Per the high net retention rate, organizations seem to be finding the resulting automation beneficial enough. Regardless of what one may think of the company, customers are expanding the volume of business they do with UIPath and so, dismissing it too quickly may be an error.
Retention rates are world class: NRR (net retention rate) was 127% (excluding foreign exchange impact) in Q1 2024, 139% in Q1 2023 and 145% in Q1 2022 (notice the deltas).
(NRR measures your ability to retain and expand customers).
While NRR has been declining since Q1 2022, it seems to track the overall pipeline slowdown I have seen in the B2B space across the board over the last year.
The high NRR is a result of the platform´s horizontality, according to management. Customers land in one division and then expand their use of UIPath to other divisions.
Per UIPath´s Q2 2022 ER, for the average customer the cost of deploying UIPath´s software $15,000-$25,000. For Palantir, the figure is about 10-20 times higher.
Certainly, automating clicks does not seem like a menace at first glance to a mean machine like Palantir, but UIPath is evolving into a more sophisticated solution.
In Q1 2024, UIPath released Clipboard AI which enables it to “intelligently transfer data between documents, spreadsheets and apps, understanding the content and automatically inserting the data into the right places.” If UIPath gets this feature right, it can unlock semantic automation. This would enable UIPath to access deep levels of nuance, positioning it to automate workflows of higher value and not just mechanic processes.
Instead of doing all the boring clicks for you, it can maybe at some point automatically convert a long email into a presentation or perfectly understand what is going on in the supply chain just by reading emails, for example. I have no idea how good Clipboard AI is or not, but UIPath does have a track record of constantly updating its platform while maintaining very high rates of retention, which speaks well to its ability to bring additional value to the table. Maybe it gets semantic automation wrong, but its willingness to iterate reflects a pattern of companies that tend to compound over time. Plus, what if it gets it right?
Ultimately, the point is that UIPath´s advantage in distribution (and solid retention) grants it a privileged position from which it can iterate its way to incremental degrees of automation over time, including the integration of third parties. In September of 2022, for instance, UIPath partnered with Snowflake to enable joint customers to “design and build workflows based on 360-degree views of trusted and accessible data on Snowflake’s platform.” UIPath funnels the data and Snowflake stores it: again, perhaps not that great a solution compared to Palantir, but once UIPath has the top of the funnel locked in it might not matter so much.
UIPath sits above the application layer, which explains part of its stickiness. Organizations do not have to worry about the inevitable changes at the application layer of their stack, because UIPath is simply looking at screens and is application agnostic. However, this is also UIPath´s seemingly major weakness: OS providers like Microsoft sit at the lowest level of the stack and thus have a hypothetical distribution advantage. This is the key unknown of the thesis for me, which I will explore in depth in Section 3.
Before I do that, allow me to take you on a brief tour of what I believe matters when evaluating a potential investment in the company.
2.0 Key Business Dynamics
As a leader in the RPA space, UIPath benefits from a rapidly strengthening flywheel with an upcoming inflection point.
UIPath´s primary competitors are Automation Anywhere and Blue Prism, which unless you are an avid user of all three, are quite hard to tell apart. Watching an interview of co-CEO (but founder) David Dines, however, I learned that, up until 2017, UIPath was far behind. It then invested heavily in sales during 2018 and 2019 and came to dominate the market. In 2020, Automation Anywhere and Blue Prism did the same, but got caught by the pandemic with their investments achieving much lower returns. Since then, UIPath has led the RPA market.
The timing was wonderful and perhaps miraculous, but no actual signal can be extracted from the event by looking at the numbers alone. However, the occurrence seems to correlate to David Dines´s philosophy and management style. Having spent plenty of time listening to him speak, I am convinced that UIPath is a personification of the way he sees life and business. In this interview (which I recommend you watch from beginning to end), he talks about the importance of being humble. When asked how UIPath got to lead the RPA space, he simply attributes it to listening to customers.
This struck me as the ever-so-elusive moment in which you get the opportunity to infer how the entire organization works with just one rich data point. It led me to understand that Dines is an excellent leader and that, unsurprisingly, UIPath is leading the RPA space. Allow me to break down this intuition further.
When looking at this business, there are endless metrics that you can fixate on, but in my mind there are two that truly matter:
Time to value: the quicker you can acquire new customers and enable them to obtain ROI, the quicker ARR (annual recurring revenue grows).
Retention rate: by retaining customers and adding new features that deliver incremental value, you can yield operating leverage and get ARR to grow non-linearly with respect to costs, thus hypothetically increasing free cash-flow (per share).
Poking around UIPath´s product, it seems that it simply does what customers want it to do. A simple and obvious statement, yet at scale a marginal difference in the ability to listen to customers and give them what they want makes for commensurate delta in time to value and retention and, therefore, in market share. When Dines talks about being humble and listening to customers, I can perceive to what extent he has irradiated this fundamental attitude throughout the organization.
Of course, barely anyone could have seen UIPath coming in 2017 and there is no telling when a close competitor may be able to pull a similar maneuver. But, UIPath has the DNA of a winning horse.
The more processes UIPath can observe and collect data on, the more processes it can eventually automate. Over time, this yields better time to value and retention, as UIPath is able to bring more value to every new client. At the limit, the compounding set of processes becomes something that does a whole of things that UIPath clients may not even have foreseen. This applies to its competitors too, but the leader in the RPA space (as is the case in many other digital domains) yields a further competitive advantage just by having access to more data.
To achieve the above flywheel, UIPath has developed an end-to-end platform–impressive, considering that they started as more of a consulting firm with some supplemental software. The platform’s two most important components, as pertains to identifying a potential inflection point, are:
Process/task mining: to figure out what can be automated.
Low-code development: to gradually democratize the platform, by giving increasing levels of agency to non technical folks.
A leap in the above two capabilities should hypothetically translate into a higher ROI for customers by yielding higher volumes of automation at a marginal cost over time. I have identified two key drivers that could over the coming years meaningfully enhance UIPath´s ability to mine tasks and democratize the platform:
In Section 1.0 I mentioned semantic automation, which would yield a form of intelligence that would enable UIPath to understand the unstructured data within documents and thus, automate work in the creative realm instead of just the mechanical realm currently.
The conversion of natural language to code via LLMs (large language models, like ChatGPT). This could radically enhance UIPath´s low-code solutions and hypothetically let anyone with zero technical background use the platform.
Semantic automation and natural language to code have a recursive relationship too. The better the software gets at understanding documents, the better it is able to correlate natural language to the actual work that is to be done and thus, the code that must be implemented. For example, there is a big gap between a software that understands that a presentation must be created:
Using a specific data set and color scheme.
To close a deal.
The former context is achievable via semantic automation: the software could pick up many datapoints from communications between the counterparts of a deal and learn over successive negotiations within a line of business what content translates tends to translate to success, thus gradually learning what code it must run in order to close a deal. The value of the output of the natural language command “make a presentation with this data” thus rises exponentially over time.
The advantage conferred just by having more data is amplified by the above two drivers. When it comes to training the AI models that power semantic automation and conversion of natural language to code, small deltas in the volume and quality of training data amount to outsized deltas in the performance of said models, thus decreasing time to value and increasing ROI. These two drivers are yet to materialize, but I get the gut feeling that UIPath can pull it off.
Further, some have asked me on Twitter whether I think LLMs can render UIPath redundant, fearing that LLMs can somehow learn to automate clicks on their own, since they exhibit emergence: they learn new skills without us teaching them explicitly, as we add more parameters to the models. The answer is no, for the following reasons.
To train any sort of AI, you need to first pick up large amounts of high quality data that the AI can learn from in the first place.
For the AI to add value to customers, it needs to have seen the scenarios that the customers face daily, so that it can then automate them.
In UIPath´s case, this translates into having to watch the screens of customers across different industries and different functions within the industry (sales, finance etc).
Thus, the key for automating clicks (and eventually semantic workloads) is to deploy an infrastructure at scale that can pick up data.
Therefore, LLMs on their own will not disrupt UIPath, but rather, a competitor that deploys a larger infrastructure and eventually can pick up more and better data may do so.
This brings me onto the next topic, which is Microsoft.
3.0 The Innovation Stack or, Dealing with Microsoft
UIPath may be in the Champions League (which is the football equivalent of being in the Major Leagues).
As I explained at the end of Section 1.0, UIPath has effectively rented a place in the Microsoft universe. Microsoft owns the operating system that UIPath scrapes and as such, the latter sits at a seemingly vulnerable competitive position. However, while I have encountered this situation many times and indeed companies in similar positions have gone bankrupt, a select few manage to not only survive, but thrive.
When studying Block (formerly Square), I came across what co-founder Jim McKelvey coined as the Innovation Stack: thousands of small and incremental innovations that amount to an imperceptible moat. This is the force that enabled Block to survive when Amazon came down to Earth to kill it and is also the force that has enabled Spotify to defeat Apple Music, even though the former has also rented a place in the Apple universe. Although I need to spend more time with UIPath to get a better sense, I suspect that it may be in this league too.
In the Q1 2022 ER call (the first ever in UIPath´s public history), here is what co-CEO David Dines replied when asked about competition with Microsoft:
Our approach is extremely difficult to replicate.
It requires a huge experience curve that we have built over the last 15 years. Our platform is a combination of UI, API and computer vision AI that is again extremely difficult to replicate and it is our secret sauce.
We consistently have proven and we have beaten on our competitors, with our technology.
I would naturally not make an investment in UIPath based on the above comments alone, but the aforementioned qualitative data points regarding distribution, iteration, and culture tell me they’re onto something. UIPath has only been public for just over eight quarters now, and I will continue looking for evidence that it is indeed yielding an Innovation Stack.
ARR has been growing rather spectacularly since Q3 2021, although again, we have only eight quarters to evaluate the company´s fundamentals. The income statement becomes less embellished as one moves towards the bottom line, where we find negative operating income. However, operating loss as a % of revenue has been declining steadily, on an annual basis. (150.44%) in FY2020, (18.16%) in FY2021, (56.04%) in FY2022 and (22.96%) in the TTM.
Throughout the above time period, R&D expenses have remained at seemingly reasonable levels, decreasing from 35%+ (with the exception of FY2021, in which R&D came in at 18.09% of revenue) to just over 26% in FY2023 and in the TTM–presumably as a measure of efficiency throughout the macro B2B sales pipeline slowdown. What seems to be dragging operating income into the red are SG&A expenses, which account for the lion’s share of operating expenses.
On a positive note, SG&A expenses are also declining as a % of revenue. They have gone from 194.20% of revenue in FY2020 to 79.91% of revenue in the TTM. Per these numbers, UIPath seems to be yielding operating leverage. It coincides with the company´s move towards becoming an end-to-end platform: it is likely the result of doing more business with its average customer and healthy retention levels. This view is supported by the rising customer count in the >$100K and >$1M ARR customers:
If you refer back to the chart showing the quarterly operating losses of the company, you may appreciate as I do that the losses are slowly starting to converge towards $0. It seems likely that if semantic automation and the conversion of natural language do materialize, the increases in operating leverage will push UIPath into GAAP profitability.
The non-GAAP numbers naturally look far more attractive. UIPath´s conversion from GAAP to non-GAAP essentially consists in adding back SBC, and it seems that the company is moderating SBC following some voluminous issuance at the height of the euphoria in FY2021. Perhaps this may assist in printing more aesthetic GAAP numbers, if the moderation continues, but the numbers are still very high and may destroy shareholder value.
Cash from operations has turned positive as of January 2023, tracking the convergence of operating loss towards $0. This is a very positive development that ultimately reflects the rising operating leverage.
At the end of Q1 2024, UIPath had $1.3B in cash and $0.469B in marketable securities, with no debt: although bear in mind that shareholders pay the price of no debt via the SBC issuance.
I see glimpses of greatness in this company, but there are currently two things that stand between me and conviction:
A clearer understanding of its competitive position, which I would ideally resolve by using the existing technologies in the market and talking to others that do.
More signs that indeed UIPath is yielding an Innovation Stack, which over time may enable it to thrive despite being a component of the Microsoft ecosystem.
The company has been public for just over eight quarters now and, thus, gaining insight with respect to these concerns will take some time.
At first glance UIPath seems like a silly click-automation company. Yet given its excellent distribution and retention capabilities, it is positioned to iterate its way toward increasingly valuable forms of automation. Although UIPath´s capacity to build ontologies is far behind that of Palantir´s, it has a clear edge in distribution which may over time equate to tough competition. UIPath´s high retention rates are premonitory.
The company is clearly in the midst of a financial inflection point, with cash from operations pivoting into the green recently. This has been brought about by rising levels of operating leverage by virtue of the platform’s horizontality, whereby existing customers continue to deploy it across different divisions within their respective organizations.
Going forward, if UIPath gets semantic automation and the conversion of natural code to language, it is likely to experience yet another inflection point. With a rather robust track record of deploying new features and maintaining high retention rates, it seems likely that UIPath will continue to succeed. Given the recent buoyancy of UIPath´s cash-flow, these two drivers may be meaningfully accretive.
Further, UIPath has a robust balance sheet with $1.5B+ in cash and equivalents and no debt. If its cash-flow profile continues to strengthen as it continues to gain operating leverage, it could meaningfully de-risk the thesis. At a price to sales ratio of 8, the market is not entirely ignorant of the potential of this company.
Until next time!
⚡ If you enjoyed the post, please feel free to share with friends, drop a like and leave me a comment.
You can also reach me at: