The need for Ai in Mobile P2P software


5 December 2016


Current mobile purchase-to-pay (P2P) software is simply not intelligent enough to cope with the demands of the market. The associated signature-based internal controls are also too bureaucratic. Duncan Jones, senior technology analyst at Forrester, tells Steve Dunkerley how artificial intelligence will work to solve these persistent problems.


You’re out of the office and an urgent purchase order request needs your approval, but multiple signatures are required before this can be done. The rationale behind this layer of control is to prevent maverick, wasteful and fraudulent activity in the P2P (purchase to pay) and T&E (travel and expenses) cycles. For Duncan Jones, this requirement is sluggish and creates a number of redundant processes.

“The signature on the transaction is the end rather than the means to an end,” the senior analyst at Forrester says. “We have this process where three people have to sign off before anyone can buy anything.”

The purchase order then needs more signatures before being released, which leads to the next hurdle – the invoice. This has to be matched with the purchase order, which requires more signatures. Jones says he does not understand how finance departments think these “rubber stamps” benefit the company.

As a qualified accountant, and a senior finance and procurement technology analyst at Forrester, Jones understands the current capability of digital technology and, at the same time, intimately grasps the lingering conventions of finance and procurement. This puts him in a great position to question the rationale for bureaucracy.

Finance directors tend to value multisignature internal controls to protect the business from non-compliant or out-of-policy spend. However, Jones says that 70% of transactions are approved via mobile apps, making these finance conventions out of touch.

“With mobile phones, managers are not really looking at what is being purchased,” Jones says. “A 5in screen can’t give them enough information to make an informed decision about the transaction. The fact that they are doing it on their mobile phone means they’re busy, not concentrating or in a rush. It’s my hypothesis that mobile approval as part of the process is seriously undermining financial control compared with what used to happen when we physically signed paper invoices.”

Swept away

According to Jones, the next Forrester ‘e-Procurement Wave’ will question buyers about what they are trying to do with their P2P processes and help ensure employees are spending the company’s money wisely.

“All these approval signatures were just a route to getting better control,” Jones says. “They’re not the goal. We’ve allowed ourselves to get into this habit where the signatures are the end in itself, and I think the technology can help us rethink that. They provide the illusion of control but don’t enable a manager to spot who is wasting money. Instead, we’ll shift management attention from chasing signatures to following up on those anomalies with value analytics.”

Vendors have designs that work the same on PC and mobile, but Jones says they’ve completely missed the point. The mobile version should work differently.

“Suppose you go into your procurement tool and type in ‘printer cartridge’,” Jones poses. “With the PC version, the software gives you 150 search results for different types of cartridges in the catalogue. Gradually, on the PC, you can filter them down, pick the ones you want and, eventually, find the printer cartridge you need. You just can’t do that on a mobile phone. You can’t give someone 100 search results and ask them to find the one that they want manually. On mobile, the software has to know which cartridge you want based on what you have bought in the past and which printer you bought. It then gives you the two you absolutely need. If you think of any good mobile app, it’s the intelligence behind the scenes that makes it valuable.”

Mobilising the purchase order approval process means automating it in the back end. If the machine can group the types of invoices and transactions that get approved by a manager 90% of the time, the system might learn what type of transactions to approve.

The machine would learn not to email a manager a request they would likely approve, based on historical patterns. The manager would also know that if the software has sent him something to look at, then there must be good reason and he needs to scrutinise it carefully.

One of the reasons why the purchase order approval process is not intelligent enough, according to Jones, is that software vendors are typically two years behind the market trends predicted by Forrester in its e-Procurement Wave.

“When I analysed the market for our 2014 Wave, I put a lot of emphasis on mobile,” Jones says. “A lot of the vendors asked, ‘Why are you focusing so much on mobile criteria as customers are not asking for it in their RFPs?’. I said, ‘It will be in a year or two and, if you are not developing the software now, you won’t be ready when customers start asking for it’. That’s exactly what happened, of course. Now, no one would buy a product if it did not have a good mobile interface.”

More intelligent T&E

Jones also sees flaws in T&E management approval processes operating in many companies. He uses an expository sales executive who is waiting to catch a flight to meet a client to close a deal. Delayed by six hours, the executive could have got on to an earlier flight for an extra £50 for a premium-economy-class seat.

“From an approval perspective,” Jones says, “this request for an upgrade would normally be rejected by finance, because it is outside policy, and policy is sacrosanct. Yet, the value of taking an earlier flight to the company could far exceed waiting for the delayed flight. By paying the £50, the executive can get to the hotel early and prepare for the meeting for the next day, which could help close a deal worth 100 times the value of that £50 investment.” 

Data analytics, according to Jones, can play an important role in building employee spending profiles for more intelligent and flexible T&E policies.

Finance directors need to know who spends the company’s money wisely versus who continually buys unjustified upgrades.

“Finance directors need to know who spends the company’s money wisely versus who continually buys unjustified upgrades,” Jones says. “The system can build up an employee’s credit rating, so if the sales executive makes a few bad judgements over time, his or her credit rating will go down.

“While human beings generally have common sense, the finance department often doesn’t because it regards the policy as god and has to enforce it. One way is to put in the rules, and another is to learn from the manager what they accept and what they don’t. If finance can change its mindset from thinking about policies being the end rather than just a means to an end, then the technology is there to support them.”

By using intelligent software to eliminate the obvious purchase order requests or build up insightful employee profiles, finance directors can have greater time to address the important requests, while travelling sales executives can focus their attention on developing new business.

A history lesson

Duncan Jones highlights what has happened in the business software industry in the past few years:

“If you are a buyer, you have to choose between a hybrid of companies, perhaps pretending to be SaaS, versus those that are truly independent SaaS natives from an architecture point of view. There are lots of independent vendors such as Coupa, Basware or Bravo Solutions a buyer has to choose from. Do they want true independence or someone not as SaaS-y as it used to be – and as the vendor would like to claim that it is?

“To me, the acid test is the commercial model. Some of my colleagues say, ‘Look at the architecture, is it multitenant? Is it single-tenant? Are there three monthly product updates, or is it a biannual update cycle?’. I look at the commercial model because if it isn’t SaaS, the culture and everything else probably isn’t SaaS either.

“If a company claims its product is SaaS, and they want you to sign up for a five-year fixed contract where you are paying at the high watermark you reach in terms of users or transactions, then that, to me, is fake SaaS and a big red flag. Start looking at other things about the company that may not be true SaaS either. That’s the big message: look at the commercial model. Is it truly flexible SaaS, or like the old model where you pay up front and can never get your money back or reduce your costs?

“The problem is that if I sign up for 12 months and next year my business shrinks – or I’m not using the product as much – do I have to pay the same as I paid last year or can I shrink my commitment down in line with my business? If I’m implementing it across a couple of business units and hope to roll out globally, do I have to pay now based on the volume I might reach when I roll it out globally? Or can I just pay for what I am implementing now and expand it overtime as the roll out proceeds?

“If you make a product decision and then look at the commercials, you are going to miss this key point. You have to look at the commercials, place some stakes in the ground and say, ‘No, these are the commercials we insist on’. If the vendor declines, then that’s another red flag. You have to do some more due diligence before you award them the contract.”

Duncan Jones contributes to Forrester’s offerings for sourcing and vendor management professionals, and for application development and delivery professionals. He is a leading expert on business application strategies and how to get the most from strategic software partners, including packaged applications, SaaS and custom development.