Author: Gloria Qiao, Founder & CEO of Trusli
Recently, we have discussed how to best begin deploying AI and technology in support of lawyers. But what about procurement? Granted, the adoption and implementation of a legal playbook benefit both in-house legal and procurement teams immensely. However, what about applying AI to procurement automation itself? Where do we start, and what are the opportunities for machine learning and automation?
A common place where many companies struggle is how to manage approval workflows. Where does a request start, and how do you ensure everyone along the chain is (informed and in agreement?) aligned? Because every company has a different process, some may want to start with finance while others want to begin with legal/procurement, and yet others may want to start with the budget owner’s group. Many organizations have found this process challenging to manage for many years, but now it seems technology may offer some solutions for this issue.
So this is great, but does it mean we should only be looking at procurement process automation? Not so fast. Our years of leading large procurement teams have taught us there are plenty of other opportunities to incorporate and benefit from the application of automation with machine learning in the process. Here are some ideas.
1. Automated historical data comparison
If I think back to the days when I was doing (many!) deals myself, I recall that once a deal hit my queue, the first thing I always did was to look at some historical data: have we bought this thing before? If so, for how much? Have the line items changed in quantity to justify the price change? Are we buying more stuff than we did before? Are we spending more money overall?
All of this became the starting point for my first negotiation of any deal. Of course, these kinds of comparisons and insights don’t just get served to me on a silver platter. I had to work, sometimes very hard, to get this information.
The first step, of course, is to find any historical data. If the company has an established procurement system such as Coupa, that involves searching through past POs to get it. Unfortunately, depending on how well the procurement system has been implemented, I may or may not find the structured data that I need. Sometimes, after some digging, and the procurement gods are smiling, I’ll find a quote in PDF format, but that first needs to be restructured and then parsed so it can be entered into a spreadsheet.
The second step is now to compare the historical data with current data. The current data can also come in different formats: a PDF in a quote, a word document with a table, or, if you are lucky and or experienced and know what to ask for: a spreadsheet.
Finally, I need to manually enter both sets of data into another spreadsheet and run some comparisons. This will often lead to an “aha” moment where I would say, look, we are buying 2x of item A, but how come we are only getting a 20% additional discount? Oh, and don’t forget that we are now spending 3x of the total money for this deal, shouldn’t that mean we should get even a bigger discount?
Later on, when I started to lead bigger teams and may not have had the time to do this analysis myself, I started to assign these tasks to junior team members. But they had to go through all the same steps. And during the process, they were also at risk of entering data incorrectly or making mistakes in their calculations.
And so the question becomes: since this is a repeatable/predefined process, why can’t it be automated? Intelligent software can go dig out the data and perform the comparisons with both speed and accuracy. And since there is very little fuzziness in both the input and output, with machine learning, we can generate insights like the ones above and then simply follow those insights to optimally start our negotiations.
2. Automated deal iteration tracking and negotiation insights
Great, so now you have presented the supplier with your first ‘asks’ by analyzing the historical data. But as we all know, negotiation is like horse-trading. You won't get everything you ask for. You win some, and you lose some. Now what?
A typical next step is to run another comparison with the first quote, and then continue this process for each round. For each iteration, you compare what has changed, where you have achieved additional savings, by how much, and where you still need more improvement. For a complex deal, you may go through this process for several rounds. Finally, when you are done, you need to perform an overall comparison with where you started and report on results and negotiation strategy.
Naturally, this entire process can also be automated. Instead of tracking each of these metrics by hand or by using a spreadsheet, an intelligent procurement system can be implemented that keeps track of each iteration during the negotiation process providing valuable insights to your negotiator along the way.
3. Autonomously producing deal summaries
As mentioned before, at the end of the negotiation cycle, procurement managers often need to report on negotiation results and strategies to ensure all the stakeholders are aligned and the deal is approved. Typically this involves building yet another spreadsheet and writing up a summary of the strategies deployed and steps the negotiator went through to achieve the final results.
Our procurement automation software has the ability to build the spreadsheets for you, automatically. Not only that, but we can also provide the basis and core elements of a deal write-up. We summarize the steps the negotiator went through along the way, quoting data and strategy step by step. We then outline the ultimate result achieved by using the chosen negotiation strategy. This won’t necessarily be an automatically generated final report and the negotiator may desire to add additional narrative to target the summary to the intended audience, but the nuts and bolts of the report are handled automatically in a pre-drafted summary, leaving the report author to focus on those elements important to highlight to their own organization.
4. Automation for market intel?
Finally, we get to an interesting but somewhat more challenging task. When you get a quote for X software, there are two kinds of market intel you may need: 1) how much do others pay for X?; and 2) if I don't use X, is there a Y software that can perform a similar functionality and if so, would that be cheaper for me?
Procurement managers engage in such market research for ‘should cost’ and RFPs for alternatives all the time. One can imagine that there is a database somewhere where you can achieve both through data mining and machine learning.
The challenge is knowing where to find all this data, and even after you find it, how to get rights to anonymize and aggregate the data so it can be shared. There are companies whose sole business model is this: to provide market intel about should cost and alternatives.
In the grand scheme, having such a database and automated comparison enables companies to be more efficient in price negotiations. It also brings more competition to the market (though suppliers may not like this as much!) by enabling greater data flow and transparency. However, because of the challenges of gathering, anonymizing, and aggregating such data, this may be a long-term project that needs more collaboration and alignment among a bigger group of companies.
Finally, there are much more beyond approval workflows when it comes to procurement workflow automation. How about turning an approved deal automatically into POs? The AI can even write out the line items of the POs for you. This is just a teaser example.
We know this all sounds fancy and fanciful, and we also know that procurement managers may ask: what if we already have some workflow management tools and our PO systems aren't the best implemented? Don't worry. Here at Trusli, we are helping procurement teams to implement intelligent automation on top of their existing systems.
We don't have a one-size-fits-all, pre-built, full-blown system that forces you to change the way you work. Instead, our approach is a no-code, drag-and-drop, modular one. Show us what you've got, and we can build on top of that. Be it a procurement system, a deal intake system, or anything else. Our approach to machine learning and data is derived from self-driving cars. We help you achieve true automation, which may start with an approval workflow dashboard, but then goes to Infinity (or even beyond!).
Sounds intriguing? Join us as our design partner. We are only looking for three. Grab a spot before they are gone. Wonder why we are the right people to do this? Check out our latest blog article here.