A decade ago, Sourcing Innovation published a piece on how Every Check Has a Cost which echoed a point made by Paul Graham that one of the big differences between big companies and startups is that big companies tend to have developed procedures to protect themselves against mistakes while a startup walks like a toddler, bashing into things and falling over all the time and, as a result, over time, gradually puts in place rules and procedures and associated checks and balances to prevent it from falling over itself, especially when the fall results in a mini-disaster (such as a contracted supplier going bankrupt).
Thus, as the company grows, it will invariably accumulate more checks, either as responses to disasters or as a result of hiring people from bigger companies who bring more checks with them for protecting against disasters which have not yet happened (and which may never happen).
But this isn’t necessarily a good thing. Unnecessary checks cost time to document,, implement, support, and maintain, especially if it’s for a situation unlikely to happen or a situation that, when it happens, will cost the company less than the cost of the check and balance it has to go through day in and day out. For example, like checking the references and solvency of an office supplies, furniture, or an off-the-shelf electronics provider. Who cares. One goes out of business, 10 more down the street.
Or mandating committee review and on-site demos for what should be a $10,000 piece of software. As described in our classic piece, the more expensive you make a sale, the more expensive that sale is going to be. If it costs a vendor $30,000 to sell you what should be a $10,000 piece of software, they’re going to charge you $50,000 — $10,000 for the software, $30,000 for the cost of sale, $5,000 for the additional support they expect, and an extra $5,000 to make up for the commissions they are losing spend all that time with you.
Similarly, it’s costly to have a manager check every purchase over $250 made by an employee just because someone decided that should be an arbitrary threshold.
Ten years ago we said review all your checks and balances and get rid of the ones that don’t make any sense or cost you more than they would save in the worst case.
But now we are saying don’t just get rid of those, get rid of ANY manual check that doesn’t add value the majority of the time — and replace it with an automated system check backed by RPA (robotic process automation) driven AI (assisted intelligence) that determines whether or not there is enough risk to warrant a manual check.
With good risk models, good training data (common situations when a “mandatory” check resulted in an approval, common situations where a “mandatory” check resulted in a denial, and exceptional situations where a “mandatory” check resulted in a request for more information by the approver), good budget/spend data, contract/catalog data (and preferred suppliers), and organizational hierarchies (with well defined roles), a system can not only easily map into (definite) yes / (definite) no / more information / forced manual review buckets and improve its knowledge of typical organizational purchase and approval patterns over time and reduce the number of manual checks to those situations that are truly risky or truly unclear. Which is, to be precise, the only time a check should be applied. (And over time, it will be able to suggest better and better check rules that help an organization understand what, and only what, it should truly be checking.)
And when you implement the right software to automate these mostly unnecessary checks (on the road to eliminating them), just like you can slowly take the foam off the table corners and the training wheels off the bike, you will grow up as a purchasing organization and, after finally finding a proper use for RPA and AI, you will say:
Domo Arigato, Mr. Roboto Patoron!