© Pobytov/Getty Images

Bots, algorithms, and the future

of the finance function
Automation and artificial intelligence are poised to reshape the finance function. Knowing what to automate
and managing the disruption can lead to a new era of productivity and performance.

Frank Plaschke, Ishaan Seth, and Rob Whiteman

For years, a global pharmaceutical company Offshoring, outsourcing, and centralization have
had outsourced its procure-to-pay finance activities, been the bread and butter of improving the finance
such as processing invoices and paying suppliers. function’s productivity for decades. As the
Savings from low-cost labor and improved processes pharmaco’s experience shows, tech-savvy CFOs
had yielded savings, but managers were eager are now considering automation to propel a
to explore whether automation could unlock new new wave of efficiency and performance. By our
opportunities. After assessing for themselves assessment, the economics of automating
how much work could be automatable, estimating many finance activities are already compelling—
the value at stake, and calculating the invest- a resounding success in some areas, even if
ment required, they challenged the company’s performance is mixed in others. Today’s cheaper,
offshore business-process outsourcer (BPO) to show better, and faster technology seems destined to
that it could compete with an automated model. reshape the finance function—and without the multi-
In the end, the pharmaco managers decided not to year headaches that many CFOs associate with
bring the outsourced elements home to automate. early enterprise-resource-planning installations.
But they did renegotiate the company’s BPO
contract, saving 40 percent or more over the next As in other business settings where automation has
three years. become increasingly viable, its implications in

Bots, algorithms, and the future of the finance function 18

 finance look to be disruptive for companies and and organizations around the technology in a fun-
outsourcers alike. The trend raises issues that exec- damental way. And they will need to manage
utives must consider as they adopt a more auto- the disruption to get through the effort without
mated finance operating model, whether internally breaking an already stretched function.
or through outsourcing. For starters, automating
the finance function
McKinsey may be
on Finance 65enticing
2017 conceptually, Understand what can be automated
but benefits
Finance can be elusive. CFOs will need a
automation Finance organizations perform a wide range of
clearer understanding
Exhibit 1 of 2 of what kinds of activities activities, from collecting basic data to making
can be automated. To take full advantage of the complex decisions and counseling business leaders.
opportunity, they’ll also need to rethink processes As a result, the potential for improving perfor-

Exhibit 1 Transactional activities are the most automatable, but opportunities exist across
most subfunctions.

Activities that can be automated using demonstrated technologies, %1

Difficult to Somewhat Highly Fully

automate automatable automatable automatable

General accounting operations 12 12 77

Cash disbursement 18 4 79

Revenue management 4 17 4 75

Financial controlling and external reporting 9 18 36 36

Tax 19 24 19 38

Financial planning and analysis 11 34 45 11

Treasury 18 43 21 18

Risk management 20 60 20

Audit 40 40 10 10

External relations 33 67

Business development 100

1 Proportion of tasks. May not add to 100% due to rounding.

Source: McKinsey analysis

19 McKinsey on Finance Number 65, January 2018

 McKinsey on Finance 65 2017
Finance automation
Exhibit 2 of 2

Exhibit 2 Many activities in the finance function can be automated.

Accounting Accounts payable Accounts receivable

• Automating complex journal • Entering nonelectronic-data- • Generating and validating invoices

entries interchange invoices • Applying cash to outstanding
• Performing and documenting • Performing 2- and/or 3-way invoice balances
account reconciliations matches • Analyzing and processing
• Calculating and applying • Processing expense-approval disputes
allocations requests • Creating reports (eg, accounts-
• Maintaining fixed-asset • Completing audits (eg, duplicate receivable aging, credit holds)
accounts supplier payments)

Financial planning and analysis Payroll Other

• Building standard • Flagging time-sheet errors and • Preparing external-reporting

management reports omissions templates
• Consolidating and validating • Auditing reported hours against • Conducting transaction audits of
budget and forecast inputs schedule high-risk areas
• Gathering and cleaning data • Calculating deductions • Preparing wire-transfer requests
for analysis • Harmonizing data across multiple
timekeeping systems

Source: McKinsey analysis

mance through automation varies across sub- and optical character-recognition tools, have
functions and requires a portfolio of technologies been applied successfully across a number of activi-
to unlock the full opportunity. Applying the ties in finance (Exhibit 2).
same methodology outlined in the McKinsey Global
Institute’s automation research, we found that Many of the technologies that enable basic task
currently demonstrated technologies can fully auto- automation, including robotic process automation,
mate 42 percent of finance activities and mostly have been around for some time—but they’ve
automate a further 19 percent (Exhibit 1). been getting better, faster, and cheaper over the
past decade. Moreover, many automation
About a third of the opportunity in finance can be platforms and providers were start-ups a decade
captured using basic task-automation technologies ago, when they struggled to survive the scrutiny
such as robotic process automation (RPA). of IT security reviews. Today, they’re well
Working atop existing IT systems, RPA is a class established, with the infrastructure, security, and
of general-purpose software often referred governance to support enterprise programs.
to as “software robotics”—not to be confused with Today’s task-automation tools are also easier to
physical robots. RPA and complementary deploy and use than first generation tech-
technologies, like business-process management nologies. Where a manager once had to wait for

Bots, algorithms, and the future of the finance function 20

 an overtasked IT team to configure a bot, today a into the existing process. Moreover, managers often
finance person can often be trained to develop see automation as a technology initiative that can
much of the RPA workflow. Today, we estimate that be led by the IT department. As a result, companies
it makes sense from a cost/benefit perspective end up with a patchwork of incongruous technology
to automate about half of the work that can be tech- tools that automate separate and distinct parts
nically automated using RPA and related task- of the process. This approach is fine for capturing
automation technologies. the first 5 percent or so of automation’s impact.
But unlocking the full potential requires a funda-
Capturing the remainder of the opportunity requires mentally different way of thinking.
advanced cognitive-automation technologies, like
machine-learning algorithms and natural-language To capture that potential, managers must be willing
tools. Although they are still in their infancy, that to reengineer their processes completely. At
doesn’t mean finance leaders should wait for them one global financial company, for example, they
to mature fully. The growth in structured data systematically went through each part of the
fueled by ERP systems, combined with the declining record-to-report process, redesigning the activities
cost of computing power, is unlocking new and organizational structures around a portfolio of
opportunities every day. technologies. These managers used task-automation
technologies like robotic process automation for
One tech company, for example, developed an purposes such as preparing journal entries, as well
algorithm that monitors internal and external data as cognitive-automation technologies like machine
to audit expense reports. The algorithm cross- learning to reconcile differences between disparate
checks them against travel data and personnel data— accounting records. Although they haven’t yet
since travel needs vary by role and rank—to begun deploying natural language tools to produce
highlight potentially fraudulent activity. In this report commentary,1 they have not only proved
case, the company uses the output to identify that these technologies work but also designed their
areas where policies may be unclear, not for enforce- processes to adopt them down the road. The
ment. A similar effort enabled the company result was a road map that these managers expect
to audit vacation time continuously: an algorithm to unlock 35 percent savings from automation
compared declared vacation days with data over the next two years.
from badge swipes and computer-usage data to
confirm whether employees were reporting At a heavy-equipment producer, managers had
vacation time accurately. Cases like these represent long used spreadsheets to forecast monthly sales
the beginning, not the end, of what’s possible and production. Frustrated with the time
with cognitive-automation technologies. consumed and the imprecision of manual forecasts,
they tasked a team of four data scientists with
Rethink people and processes around developing an algorithm that would automate the
the technology entire process. Their initial algorithm used
Today, processes in the finance function are all the original sales and operations data, as well as
purposefully designed to harness the collective additional external information (about weather
brain power and knowledge of many people. and commodities, for example). In this case, within
The temptation for managers as they implement six months the company eliminated most
an automation program is to follow that same of the manual work required for planning and
pattern, retrofitting a particular automation tool forecasting—with the added benefit that the

21 McKinsey on Finance Number 65, January 2018

algorithm was better at predicting market changes don’t fill existing roles as people leave. Also, such
and business-cycle shifts. roles usually don’t require a major organiza-
tional redesign to capture automation’s benefits.
Manage the disruption A team that currently requires 20 people could
In theory, finance has many opportunities to simply reduce its head count to ten by using a fully
redeploy its people. Financial-planning and -analysis or partially-automated solution. Going after
professionals could be retasked to support basic tasks first allows the remaining employees
the business. Tax specialists could be refocused to focus on the more professionally rewarding
to maximize after-tax income. tasks, and early wins create the capacity and fund-
ing that help the finance function to fund other
But, especially in transactional functions, the parts of the automation journey by itself.
hard reality is that automation—if implemented
effectively—will inevitably lead to changes in One institution started by rolling out some
organizational structures, redefined roles, and 200 bots to automate work at its offshore shared-
layoffs. At one global financial institution, the service centers. That allowed the company
CFO is on pace to release a quarter of the company’s to develop a playbook, a governance model, and a
20,000-person shared-services organization workforce-management strategy that could be
over the next 24 months. That’s bound to be disrup- deployed elsewhere. It also created the foundation
tive, and there’s no point in pretending these needed to consider automating more complex,
realities don’t exist or trying to hide an automation higher-order processes, such as financial
program behind closed doors. modeling and audit.

The leadership and vision of the CFO, in ƒƒ Create a human-resources and placement
particular, are paramount, just as with any finance capability that works in lockstep with the CFO and
transformation. In our experience, the best the finance function. Automating more complex
approach is to manage automation systematically activities, such as a company’s controllership and
along these lines: tax functions, often means releasing people,
since these have less turnover than more transac-
ƒƒ Start with the more mundane, transactional tasks, tional work. For many companies, redeploying
which inherently have higher turnover. Rather than people has proved a challenge. Most just take the
releasing a lot of people, in many cases you just savings or, worse, incur new automation costs

Especially in transactional functions, the hard reality is

that automation—if implemented effectively—will inevitably
lead to changes in organizational structures, redefined
roles, and layoffs.

Bots, algorithms, and the future of the finance function 22

 without a corresponding reduction in labor spend- used to automate activities ranging from financial
ing. Thoughtful workforce planning is critical. forecasting to internal audit. Each year, two
or three students from a local university spend
Communicating a plan for the affected workers the summer building algorithms and bots.
well before automation tools are introduced Not all of these efforts succeed, but the company
can help. The necessary steps include designing has begun implementing at least half a dozen
the future organizational structures, telling solutions developed by the interns. Similar pro-
people exactly what you’ll do to evaluate them grams will be critical to attracting talent
fairly, and promising to do your utmost to that can lead an increasingly automated
create opportunities for redeploying personnel. finance function.
Maintaining a constant lineup of open posi-
tions in finance and other parts of the company
can further minimize the impact on people.
Honesty and transparency are critical. Automation is already reshaping the future of work
in the finance function, and the opportunity
One North American bank, for example, explicitly to boost performance will fuel the trend. Adapting
mapped the automation solutions it was using to to disruption is challenging, but CFOs who
the approximately 200 finance employees affected. build a clear early perspective on the nuances of
Before the organization introduced the tech- the automation journey will be well positioned
nology, it had a plan to redeploy employees in more to thrive.
valuable roles. To date, the company has found
ways to redeploy nearly 50 of them to other areas
within and outside the finance function. 1 As opposed to commentary written by people.

ƒƒ Adapt the recruiting and retention profile to get Frank Plaschke ([email protected])
the finance professionals you need. Even if is a partner in McKinsey’s Munich office, Ishaan Seth
technology intimidates some employees, a willing- ([email protected]) is a senior partner in the
ness—and ability—to learn new tools is impor- New York office, and Rob Whiteman (Rob_Whiteman@
tant. Future leaders will be quite excited by McKinsey.com) is a partner in the Chicago office.

a function on the leading edge of digitization and

Copyright © 2018 McKinsey & Company.
automation. And even CFOs of companies
All rights reserved.
that aren’t planning an automation program in
the next year or two should seek out and
recruit people who will be prepared for it when
it happens.

One technology company undertook such an effort

by creating an internship program to attract
machine-learning talent to the finance function.
The company maintains data sets that can be

23 McKinsey on Finance Number 65, January 2018