Alan Holland, Founder and CEO of Keelvar, AI-enabled Sourcing Automation & Optimization Software for enterprise procurement teams.
During a conversation with ProcureTech’s CEO Lance Younger, Alan shared his thoughts on scaling sourcing excellence, agents of change, economic injustices, algorithmic mechanism design, moral obligations…Playing Bigger, Strava and Murphy’s.
1. What is Keelvar’s mission and what do you see as being the key differentiation for you?
Our mission is to help procurement achieve excellence in the sourcing process but also, to help suppliers engaging with the buying community express their strengths and weaknesses.
2. How has that mission evolved over the last couple of years?
It’s been a 2-stage process.
First, we built a sourcing optimizer: it’s an advanced sourcing tool that allows big businesses to facilitate very expressive communications with their suppliers, whilst being very expressive about how they want to balance different trade-off costs and non-cost objectives. This tool allows companies to achieve excellence in their negotiation processes.
Stage 2 involved building software agents/bots that automated negotiation actions. When relating this to our mission, we’re talking about scaling sourcing excellence.
3. What have been the most significant decisions in the Keelvar journey so far?
We were reaching a stage of feature completeness for our source optimizer. Then we started embarking on the second part which was very much untested. We couldn’t say for sure if the market would adopt sourcing bots or whether it might be seen as a step too far: you have to take a leap of faith.
When we first started building sourcing bots, we had to convince everybody within the company that this is what the future would look like. We were looking at what we were working on in the university research lab before we set up Keelvar and we used to have trading agent contests; different university research labs would compete against each other and trade in this contest where it simulated a year in the life of a laptop manufacturing company, and you would have sourcing bots, sales bots and scheduling bots, inventory control: everything was fully automated and fully autonomous right from when you pressed go on day one of 365 days. The contest would simulate market conditions like changing prices or demand and your bots had to react, and coordinate. Not only did they have to coordinate between themselves, they had to coordinate with suppliers. Throughout the course of the contest over several years, we saw these bots evolving along with different techniques from AI and the game theory and mechanism design through to optimization reinforcement learning and artificial neural networks.
4. What do you look for in the perfect customer?
We had 50 or so blue-chip enterprises as customers of our sourcing optimizer when we started building our Automation offering. I think that they were so keen that it kind of consumed our capacity pretty quickly. We focused on an existing customer base for automation and what we saw within those organisations, was that they were now starting to see the application of optimization to a much wider variety of sourcing categories and a huge scale up in the number of sourcing events that would pass through the optimizer.
For example, a big enterprise that reserves sourcing optimization for their top 20 sourcing projects, now they could be looking at thousands of sourcing events or even tens of thousands of sourcing events, that could be below 50k or even 10k in spend. You can still see the advantages and benefits of applying optimization at this level. Now it’s become achievable to apply optimisation at that level.
5. What are the foundations of a great team? How have you gone about putting in place a great team to be able to enable you to scale?
When we were looking at automation taking off in early 2020, we were a team of about 20 people, we were quite small and largely made of science and engineering graduates. We’re techies and we needed a strong commercial team to complement what we were doing, as well as a customer success team and marketing team. On our Board, we are very fortunate to have the former CEO of CombineNet, Rich Wilson who’s an industry veteran, he’s a fantastic mentor for us and has helped us hire more amazing people.
When competing sales people see that we’re winning deals, it becomes a lot easier to recruit. We’re now up to 60 people so you know within 12 months we’ve gone from 20 to 60 and probably heading towards 100 within the next six months.
6. What are the foundations of a great team? Are there any key values?
We are very transparent, structured, everybody’s accessible and we share a lot of information internally about what’s working and what’s not working.
Fairness, and being just, it pervades our culture, because when we started out, the reason we started tackling sourcing optimization in the first place is that we saw how core procurement processes are unjust for suppliers. Especially smaller suppliers who find it difficult to compete for business against big enterprises, because they can’t shape the nature of the contracts they need and can’t express the capacity constraints and so on.
My first experience of it was when I was helping up my parents’ company, a small chemical supplier, and they needed to be quite descriptive about the combination of locations that wanted to win, including types of volumes of chemicals they need to supply in order to be efficient and meet the needs of suppliers. It’s a kind of existential crisis for suppliers, if they can’t or if they are forced to bid on lots that are too big or they don’t know what combinations of smaller lots they’re going to win.
That’s what we would like to pervade through the culture of the company. That we treat our employees extremely well and everybody knows what’s going on.
7. As we emerge from COVID, what will you be doing differently?
We certainly will be operating differently, me and my team.
I think there’s a good opportunity for the world to reduce unnecessary travel: there are some journeys which have a meaningful purpose which is amazing, but it’s also a good excuse to explain to some of our biggest customers as well that we don’t necessarily need to be travelling to Paris or New York or Lisbon, or wherever else for some meetings that are an hour long, it’s not very sustainable nor very environmentally friendly. It’s not an efficient use of time either.
And we’re looking at different business functions. There’s some that need to meet up on a more regular basis and others that can do more of their work remotely, and we’re trying to maximise remote working side of things.
8. What is the vision for Keelvar in the future and what will be most important to achieve it? What does great look like in five years?
That’s a tough question. Actually, it’s difficult to see where we’ll be in 12 months’ time, we’ve tripled in size in only a year! I don’t know if we have to maintain that level of exponential growth but we are certainly going to be growing strongly over the next five years.
We view sourcing as a type of game.
I used to teach game theory in university, it is a game of incomplete information which is not too dissimilar to poker, because you have expectations about what other people have in their hands. You have probability distributions over what you think they might have, given the knowledge you’ve got about what’s before you. Likewise in sourcing it’s the same.
When I worked as a researcher in an AI laboratory, there was this debate whether a poker bot would ever defeat the world champion which we’d seen happen in other games. However, many AI researchers felt that it was too big a challenge for AI and lo and behold, about five years ago a researcher from Carnegie Mellon University led a team that built a bot to beat the world poker champion. The researcher, Tuomas Sandholm created the first software company, CombineNet, to build a sourcing optimizer.
So there’s a deep connection between sourcing optimization, and what we know as ‘algorithmic mechanism design’. This is an inverse to game theory on how you design a game so that, if self-interested agents are the actors involved in that game, then you can maximise the value to the person running the game, similar to sourcing bots so you can view the sourcing processes again. You can apply algorithmic mechanism design to negotiation, so you can start to learn over time what negotiation tactics work best. The sourcing optimizer has all the parameters you can control; these are the levers in the game and you need to have sourcing optimization as the underlying engine… as the game controller… because you need precision and nuance to control a mechanism in fine detail. In sourcing, details really matter.
In algorithmic mechanism design, you can tweak those parameters. If your ocean freight sourcing bot has started to learn intra-Asia routes which have stiff competition, you can change your mechanism to run a faster mechanism to play different types of feedback.
It may make sense to do your big annual strategic negotiations manually because there are aspects of trust and communications between humans that machines will ever be able to replicate.
Ultimately, we’ll have sourcing bots that are the best buyers in a range of different categories that will exceed any human expert performance. So it becomes a de facto norm, just as if you’re entering a poker contest, and it was online and you had use of a poker bot, you felt that you can win the contest. Likewise, in sourcing business decisions, you’re going to gravitate towards whatever the most effective strategy is.
In five years’ time, we would see that sourcing bots become very commonplace and the de facto norm, and how you do tactical buying.
We’ll also see automation on the sell side. As I was saying, you know the trading agent contest, getting back to the laboratory experiments we did in university days, there were also bidder bots. So, we’ll be providing infrastructure to sales teams to automate responses to automate negotiations. What you will have are bidder robots that are participating in multi round negotiations that are fully automated. Imagine automating a 20 round combinatorial auction protocol in under 10 seconds, where a dozen bidder bots are all competing against each other and the bid robots understand the combinatorial aspects of finding efficiencies and their supply networks, storage and manufacturing processes. We know it’s scientifically achievable we’ve been there and done that in the research lab, it is more the engineering challenges around gathering the data and mapping to taxonomies, and all of the other good scientific challenges that need to be tackled, they are non trivial problems, but the scientific solutions are already in public domain.
9. How are you doing good for the planet?
We’re aware of our own kind of moral obligations. There’s many billions, in the order of $100 billion flowing across our systems, and what we’re doing is we’re automating the ability for companies to reason about environmental objectives.
We’ve connected with third party data vendors such as sea routes, linescape and Xeneta to track what is being transported, as well as accompanying CO2 emissions.
You can look at the joint optimum and consider what you’re willing to pay vs the $/kg of CO2 saved. It’s about trying to automate your alignment between corporate social responsibility objectors and environmental sustainability. It’s just challenging as we face automation with reasoning and a flood of ‘cheap’ data, but that’s where intelligent automation really delivers.
INSTANT ALAN INSPIRATION
1. What is your favourite book or blog and why?
Recently, I read Play Bigger by Al Ramadan and three other authors. It’s all about defining new categories, and I think that was quite inspirational and because it pointed to lessons learned in other domains from Dyson and through to others like McDonald’s and others outside of the tech space. How they set themselves apart and identify what is a new category and became category leaders… that was a very good read.
I’ve also been getting back into reading more fiction, which is different to when you’re running a business and reading more non-fiction and business books.
In my younger days, I was an avid reader and I have kids now… now I’m getting them to read books too. The Curious Incident of the Dog in the Night Time is a brilliant book and my kids love reading that now — Alice in Wonderland too.
2. Who is your favourite inspirational leader & why?
Sonia O’Sullivan was Ireland’s leading female athlete. She always displayed quiet determination and excellence. Her actions always spoke louder than her words and there were occasions when she was deprived of medals because of competitors ill deeds. She never complained even when she knew in her heart and the public knew she had been robbed. Such magnanimity is rare and very admirable.
3. What is your favourite app (not your own) or piece of technology?
I like fitness tracker apps like Strava; I suppose I’m someone who likes to measure things. If I’ve put in a bunch of effort and training into runs, I like to track to see if I’m improving or not.
I also like Slack, which acts as a means of communicating with the team faster.
4. What’s your favourite way to celebrate a success?
This is a tricky one, much of our success has been happening post COVID. We’ve landed a big series of funding and been hiring more staff; two thirds have been hired since COVID!
We’ve had lots of wins with new blue chips, but it’s a shame we can’t celebrate, but hopefully we can all do something together in future as the team is spread everywhere.
It’s a different way of working, for sure.
5. What is your favourite cocktail?
I’m not a cocktail person, but I’d recommend the local brew here Murphy’s, it’s a stout made in Cork! I’d recommend it to anyone, if you see it in the pub try it!
KEY TAKEAWAYS
1. Optimisation and automation enables sourcing excellence at scale.
2. Successful sales growth fuels team growth.
3. Experience enables you recruit an amazing team.
4. Be masterful in your strategy and find creative ways of leveraging strengths.
5. Sourcing optimisation addresses economic injustice and sustainability equitably.
6. Adopt new approaches to information sharing.
7. Be aware of your own procurement moral obligations.
About Keelvar
Sourcing Optimization and Automation Software for enterprise procurement teams. Keelvar emerged from a leading AI research laboratory and is the category leader in best practice technology for advanced eSourcing. Keelvar proudly serves customers such as Siemens, BMW, Coca-Cola and McKesson.
About ProcureTechSTARS
Our industry is moving forward faster and faster, empowered by innovative, progressive digital procurement solutions created and led by inspiring teams. ProcureTechSTARS are the digital procurement company CEOs and Founders that are leading the change, they are entrepreneurs, engineers and architects collaborating to transform procurement and the enterprise. In an open conversation with these leaders Lance Younger will be discussing the highs and lows of building the future now, the challenges they’ve faced, their perspective on accelerators and hot topics, and what keeps getting them up in the morning (and keeps them up at night).
