Author Archives: thedoctor

It’s Hard to Find Fraud in Big Spend Stacks …

Let’s start with T&E spend. While most organizations might believe that this spend, which is primarily for low value amounts on fairly well understood products and services, does not hide much in the way of fraud, that’s not always the case. Nor is the fraud limited to employees upgrading to business class, upgrading from rooms to suites, and spending a bit too much on drinks at the client dinner. (But even this can be very expensive. If this off-policy spend results in negotiated volume-based rebates failing to materialize, this can be very costly.) But that’s not the case. It cal also contain:

  • the same receipt for a $500 business entertainment submitted two (three, or even five) times, one month apart, on different claims and never noticed
  • a pet hosteling bill that looks just like a hotel bill
  • an invoice from Benny’s buddy Bob for 20% above market rates who drove him to the airport (instead of a licensed service at market rates)
  • that double billing by your no-longer favourite hotel for a room charged to your guest and then charged on your tab is really hard to spot (especially when some rooms were picked up and some rooms weren’t at your recent event)
  • collusion between an employee and a spouse who owns a travel “services” company can account for a lot of extra travel “services” billings that weren’t delivered
  • suppliers who know you have holes in your T&E monitoring can submit fake invoices for services never delivered
  • etc.

It’s really hard to find these low-impact fraud needles in a T&E haystack, but these needles can add up quickly — especially for products and services never even delivered! Only automated processing that can compare multiple entries across multiple dimensions and learn typical patterns can identify the majority of errant fraud that passes through your T&E system.

Moreover, as an organization learns to detect certain types of fraud, the fraudsters get smarter. No static system can keep up! AI based systems are key to an organization’s success.

In particular, AI-based systems that can work on multiple types of spend. T&E is just one category. There’s also invoice data for sourced and procured products and services that can be six to eight times the T&E volume in an average organization. And when we go broad, there are even more options for creative fraud from less-than-honourable parties. For example, you could see things like:

  • $4.95K shipping fees for $5 items because the tolerances in the system don’t kick anything up for review with shipping less than $5K
  • invoices from fake suppliers with the same name as your tendered suppliers with faked registry numbers and different bank information for payment
  • invoices from corporates owned by spouses of employees for services not delivered submitted by the employees and approved by colluding associates doing the same thing
  • etc.

For some of these instances, humans have almost zero chance of surfacing the infraction when its 1 invoice in 1000. A new solution is needed. A number of players are tackling the problem with modern AI solutions, but do the approaches have what it takes to find the gold in them there hills? Only time will tell.

Sourcing the Day After Tomorrow … Part XVI

In this series we have been reviewing sourcing today, the primary phases and sub-steps, and how they look strategic on the surface but often hide a lot of tactical work underneath. Moreover, sometimes “strategic” is simply a decision that is entirely based on the results of a sophisticated analysis that can be encoded in a very complex rule.

What does all this mean? It means that systems can do more of the work and with next generation sourcing systems, the strategic decisions will be made by expert buyers who know the market in ways designers of systems can’t. Expert buyers who can identify external stimuli that occur, and impact, the market once every five to ten years (that a new system wouldn’t know). Expert buyers who can better judge the impact of a new supplier on the market that the system doesn’t have the history on. Expert buyers who know the best way to handle unexpected demands or change requests in a negotiation process.

Strategic will change from data gathering to data analysis to knowledge evaluation where the analyst first learns to analyze the data gathered to better train and correct the system to knowledge evaluation where the analyst learns to identify the gaps in the analysis or the weightings that need to change. It’s going to become primarily an intelligence exercise, not an analysis exercise. Computers can do considerably more analysis and number crunching than we can in an exponentially smaller amount of time. As a result, more and more analysis will be given to the computers, and more and more intelligence will be expected of the user.

And the entire sourcing process will be affect. How much? In the beginning, more and more of each step, and then of each phase will be automated. But then, in the longer term, the sourcing process will change and adapt to one that is more suitable for the knowledge-based endeavour that it is. What will this look like? Time will tell, but we have our ideas. And we will address them in at a future time.

Sourcing the Day After Tomorrow Part XV

In this series we are doing a deep dive into the sourcing process today, and, in particular discussing what is involved, what is typically done (manually), and whether or not it should be that way. We have already completed our initial discussion of the initial project request review phase, the follow up needs assessment, the strategy selection phase, the communication phase, the analysis phase, and the negotiations phase. Now we are in the final contracting phase. At first glance, it looks like this is the second most strategic and human-driven phase there is, second only to negotiation, as it is humans (and lawyers in particular) who typically define standard terms and conditions, humans who identify risk and mitigation strategies, humans who define obligations, and humans who analyze the contract for compliance to goals. But is this the case?

So in this final step, the contract step, we have these final sub-steps:

  • Standard Terms and Conditions
  • Modification & Risk Mitigation to Supplier & Country
  • Key Metadata definition and obligation specification
  • Contract Analytics

If all of the standard terms and conditions are in existing contracts and the contract clause / template repository, there’s no reason that a system cannot automatically scan the contracts and repositories, identify the standard organizational terms in every contract, identify the standard terms for the category, and identify any terms, often not included, that would be relevant to the category. Probabilities can be applied and contract terms organized by weight. The buyer can then just bulk select or bulk reject the relevant clauses.

In the modification and risk mitigation step, a contract analytics engine can be applied to determine how well a particular clause addresses a certain risk of relevance to the organization based on context models and differentials. It can then compare that clause to the clauses that best address the risk and identify the necessary modifications, and do so specifically from a supplier or geographic context.

In the key metadata definition and obligation specification step, the goal is to identify the right metadata that needs to be tracked against the contract. This will be dependent on the terms and conditions, the goals, the obligations, and other key information that will be specific to the contract. However, contract analytics can identify, or at least suggest, much of this as well automatically based upon similar contracts, similar terms, similar goals, and similar obligations. This can greatly reduce the effort required by a buyer.

In the final step, the contract analytics step, the identification of risks, variances from a norm, and non-standard clauses can often be better identified by a contracts analytics engine that can cross-compare potentially risky clauses and variant clauses across hundreds, if not thousands, of contracts and identify deviations from the norm. A user just has to decide whether the variance is enough to be of interest to them, and properly setting a threshold can eliminate the majority of those variances that are not.

In other words, at the end of the day, contract analytics identifies the majority of standard terms and conditions that are of interest, the majority of standard clauses that will need modifications to address supplier and country risk, the relevant metadata and obligations associated with the contract, and any clauses that can be considered variant enough to warrant special consideration.

The majority of the work can be automated with a good contract analytics engine — the role of the buyer is to apply their intelligence to determine how accurate and effective it is. As the buyer trains the engine, it will become more and more accurate over time and the strategic work will be reduced to hours, sometimes minutes for simple contracts, compared to days or weeks.

In other words, the more we explore the sourcing process, the more we find out how truly tactical, or at least automatable, the majority of it is.

Sourcing the Day After Tomorrow Part XIV

In our series to date we have recapped Sourcing today and taken a deep dive into the key requirements of the review, needs assessment, strategy selection, communication, analysis, and negotiation phases. In each of these six steps to date, we found that while some steps were critical for a sourcing professional to undertake, others, while necessary, were a complete waste of skilled talent time as the majority of the tasks could be automated. And while we’re still at the point where some tasks have to be done by humans whereas no matter what, we’re almost certain that this is true across the entire sourcing cycle, but until we complete our analysis, we can’t be 100%, so that is what we’re going to do today and tomorrow.

So in this final step, the contract step, we have these final sub-steps:

  • Standard Terms and Conditions
  • Modification & Risk Mitigation to Supplier & Country
  • Key Metadata definition and obligation specification
  • Contract Analytics

In the standard terms and conditions step, the buyer identifies all of the organizational standard terms and conditions that are relevant to the product and services in question. This involves reviewing the standard conditions proffered by legal, previous contracts, and standard contracts put forward by competitors and selecting those that are relevant.

In the modification and risk mitigation phase, the buyer identifies which standard terms and conditions, prior contracts, and suggested terms (defined during the early phases) need to be modified to address risk on a supplier and/or country basis and makes some suggestions as to what needs to be done.

In the key metadata definition and obligation specification phase, the buyer needs to define the metadata that needs to be tracked against the contract, how it needs to be tracked, where it needs to be used, and even how to generate value from the metadata.

Finally, the user needs to analyze the contract for risks, variances, and clauses that are non-standard, identify, catalog, and track them over time. Plus, the user needs to determine the relative risks, variances, and clauses relative to other contracts to determine overall priority.

This sounds pretty buyer intensive and strategic, right? Not much room for automation, right? Well, we’ll find out in our next part!

Sourcing the Day After Tomorrow Part XIII

In this series we are doing a deep dive into the sourcing process today, and, in particular discussing what is involved, what is typically done (manually), and whether or not it should be that way. We have already completed our initial discussion of the initial project request review phase, the follow up needs assessment, the strategy selection phase, the communication phase, and the analysis phase. Now we are in the negotiations phase. At first glance, it looks like this is the most strategic and human-driven phase there is — as it is us who do the negotiations, figure out our BATNA (best alternative to negotiated agreement), and determine what facts we will use in our negotiations, but we have been fooled before.

We are now discussing the negotiations step, which has the following steps that have to be completed every time (and not just sometimes):

  • Format Selection (online, offline, hybrid)
  • Fact Prep
  • BATNA fallback
  • Audit Trails

Let’s start with format selection. Sure, it’s the buyer who selects the format, but, like strategy selection, the selection of negotiation format also depends on should cost analysis, market costs, supply vs. demand market trends, and previous performance of options in similar situations — all of which could have changed since the initial event was kicked-off. Depending on the expected savings or value expected, it may not be worth the in person negotiations. And who’s better at computing the costs, computing the trends, computing the variance of current supply market context against previous contexts, extracting the differential savings between contexts, and generally at doing hundreds, thousands, and millions of calculations. The machine. In this phase, the platform could do all of these calculations, apply a few probabilistic models, and come up with a ranked list of the best options under a well-defined set of assumptions. Most of the time, especially when market costs and trends change slowly, the buyer will be able to review the options, validate the assumptions, and choose one of the best options and have the system automatically generate a report that validates their format selection. It’s a strategic human decision, but one that can often only take a few minutes after the machine takes days (or weeks) of work away.

Now let’s move onto fact prep. In this phase, once the senior buyer has selected the format for the negotiation, and revised their expectations, they need to gather all of the facts in one place that they expect will assist them in their negotiations. Besides deciding what they need, this is a very tactical phase of information gathering and consolidation — which is something the machine is best suited for. In addition, based on all of the decisions made to date, if previous events were captured as well as materials selected and used, the machine can also apply probabilistic models in this step to determine which facts will likely be most useful to the buyer and auto-generate a suggested “fact-book” (outline) that the buyer can update with minimal effort. Then, with one press of the button, all of the information they want in the negotiations is at their fingertips.

Before negotiations actually begin, the buyer will finalize their BATNA. While the best buyers will actually start outlining this during the strategy selection phase (as it will need to be executed as soon as the strategy fails, which typically won’t be until negotiations, but if the event tanks in the communication phase (not enough suppliers respond to the RFQ, prices don’t decrease from initial bids in the auction, etc., it may be sooner — and if its sooner, the phases between failure and BATNA get skipped), they won’t finish until just before the first volley of negotiations get underway (as market dynamics can change significantly between the start of a complex project and the negotiations, even with a lot of machine assistance, because the need to involve a lot of stakeholders can draw an event out and the reality that an unexpected mine or factory closure can happen at any time can flip market dynamics on a dime).

So how does one determine a BATNA? One way is to select the next best strategy (extend the current agreement, spot buy — possibly with an auction, use an alternative product design that would allow for a new event, etc.), and as we know from Part VII, the machine can help greatly in this step as it is capturing all the knowledge to run probabilistic models to rank the next-best alternatives under current assumptions.

And last and not least we have unalterable, secure, always queryable audit trails. We all know most modern enterprise systems were made for this. Nuff’ said.

In other words, the more we explore the sourcing process, the more we find out how truly tactical, or at least automatable, the majority of it is. But we’re still not done, so in our next two parts we will explore the last phase — creating and signing the contract.