Weekly Briefing

article sponsor image
Partner Content

Solving the Tribal Knowledge Gap with AI: 3 Lessons from a Public REIT

December 15, 2022 | David Stifter, Managing Director, DigitalBridge & Founder, PredictAP
5 min read
listen to article Listen to this article

It’s the end of the month. Your accounting team is fielding calls from vendors about unpaid bills while juggling monthly reporting and reconciliations. Everyone is stressed. You just hired two new AP specialists a few weeks ago, but the backlog of unpaid invoices isn’t shrinking. It’s no better than the month before. How could this be happening again?

Rising turnover in accounts payable jobs is compounded by difficulty hiring and onboarding new talent into roles known for being heavy on manual data entry and light on career potential. Real estate investment and property management companies got used to hiring more staff to accommodate the extra work that accompanies growth, but now many struggle with recruitment and retention.


Staffing aside, more organizations are coming to the conclusion we reached at hiring doesn’t easily address the essential ingredient in a well-run AP operation – tribal knowledge. We set out to solve this problem at a large public REIT and succeeded. Here’s what we learned along the way.


The Cost of AP Inefficiency

Anyone familiar with the stressful scenario above knows the true cost of inefficient AP operations: missed and late payments lead to unhappy vendors, late fees, and poor D&B ratings that can impact future cash flow; spikes in invoice volume force rushed processing, leaving businesses more vulnerable to fraud attempts and accidental overpayment.


Even after implementing AP automation to streamline AP workflows, bottlenecks persist alongside paper-first manual processes, limiting the ROI of substantial tech investments.


Containing the costs of AP operations was a priority, and the initial implementation of Yardi Payscan to automate AP approval workflows presented a relatively quick and substantial win for efficiency, visibility, and timeliness. But to fully eliminate the bottlenecks in our process, we had to solve for one key variable: our most experienced AP person-let’s call her Jane.


Unblocking the Tribal Knowledge Bottleneck

We had a Jane who had been with the firm for decades and over those years she became the wiki for the rest of the team letting people know how to break apart any invoice 15 ways – between properties, entities, deals, expense accounts, segments – ensuring it would get paid correctly and on time. This demands insights that don’t show up on the printed invoice, and take years to develop.


As we grew year over year, so did the volume and complexity of AP work. We saw a cycle of taking months to fill open positions, more months if not years to fully train up new staff and then having them leave and take that knowledge with them which started the recruiting cycle again. This was with Jane and a few long term people picking up the slack, what would happen if she retired? Would bills still get paid without her expertise? Could we even train new staff without her?


We needed a way to scale what AP could do, but the market offered no help. Most of our options were some combination of outsourcing manual data entry and using OCR, templates, or robotic process automation – all of which addressed invoice data ingestion and indexing to some degree, often with tradeoffs. None captured or codified tribal knowledge.


Watching Jane’s work more closely helped us understand that every invoice coding decision she made in a predictable, consistent way could function as an annotation, a rule inclusive of context and nuance that could be learned. That’s what told us AI, and more specifically, machine learning, could capture and apply that knowledge the way Jane could.


We built a custom solution that used past invoice history and machine learning to identify invoice coding patterns, and apply them to new invoices – like Jane, but at scale, without slow manual data entry, and decoupled from individual knowledge.


This led to massive capacity gains. Within a few months, work that once required nearly 30 full-time AP staff could be done with a team one-third the size. This meant opening up their time for learning new skills, moving them into more advanced accounting work, and increased job satisfaction due to less repetitive work. It also made us less vulnerable to turnover, and improved onboarding and productivity for new employees.


Our solution became the seed that eventually grew into PredictAP – a cloud-based invoice capture platform that could automate invoice ingestion and coding with AI.


3 Lessons for Successfully Leveraging AI and Automation

  1. Understand the Real Problem: Don’t waste precious months on misguided projects until you know what needs fixing. It’s tempting to assume AI or automation will somehow save the day, but you don’t have an AI problem. You have an efficiency or visibility or capacity problem. We had siloed tribal knowledge that caused bottlenecks and couldn’t be applied in a scalable way. Once we articulated what needed fixing, it became much easier to see what would and would not work.

  2. Clearly Define Desired Outcomes: Up-and-coming real estate CFOs and COOs are getting a mandate to pursue AI and automation, but they’re not always given a clear vision for success. For AP, the problem areas were easy to identify and quantify, once looking in the right place. Consider these examples when identifying your requirements for a solution:
    • Increase D&B ratings by paying more invoices on time
    • Realize vendor discounts from early payments
    • Shift some accounting work into accounts payable
    • Incorporate volume from new portfolios without adding headcount

  3. Examine Your Process Before Selecting Tech: A tech vendor can build almost anything, and many are incentivized to provide endless customization that keeps you coming back and boosting their revenue. The way your back office works can be highly individualized, but is the complexity necessary to preserve? The same but digital is not better, and you could inadvertently replicate inefficiencies and further entrench them into your tech stack. Identify where processes evolved around tech or staffing limitations and how isolated different steps are. This will help with problem definition, as well as make it easier to gauge whether a custom platform solution is best, or a targeted best-of-breed option will have the greatest impact.

As you consider the potential benefits of process automation and artificial intelligence in your organization, focus on the basics. Staying close to the core problem will help you assess and articulate your needs, and deliver better outcomes without having to become an expert in all emerging technologies.

David Stifter, Managing Director, DigitalBridge & Founder, PredictAP
David Stifter has spent over two decades at the nexus of real estate, technology, and finance. In 2020, he founded PredictAP to address a gap in the real estate AP automation market with an AI-powered invoice capture solution. He previously served as Managing Director at Digital Bridge (formerly Colony Capital), overseeing data architecture, process improvement, and accounting and finance transition projects for major acquisitions.

This Week’s Sponsor

PredictAP is a cloud-based invoice capture solution for real estate AP that automates invoice ingestion and coding using AI. PredictAP integrates with existing AP automation to eliminate manual data entry, improving speed, increasing capacity, and reclaiming time for high value work. Visit www.predictap.com to learn more.