How Zendesk scaled accounting across 25+ legal entities without adding headcount

About Zendesk

Zendesk is a global leader in customer experience software. Its AI-first service platform helps businesses manage customer support across email, chat, phone, social media, and other channels.

The company generates more than $2 billion in revenue, operates across more than 25 legal entities, and continues to expand through an active M&A program. That scale also demands public-company controls, even after going private.  

“We were a public company, we went private, but we never let go of the control ecosystem at Zendesk,” says Akshaya Murthy, Director of AI Transformation and Internal Audit. “We’ve been maintaining controls like we were a public company, doing certifications on a quarterly basis, with our external auditors signing off.”

For Lindsay Kelly, Director of Global Accounting and a 13-year Zendesk veteran who oversees the consolidated close, the mandate was clear: preserve accuracy and public-company controls as the business grows, without expanding the accounting team at the same rate.

The challenge

A best-in-breed finance stack created fragmented accounting work

Zendesk runs Workday for HRIS, Coupa for procurement, Zuora for billing, NetSuite as its ERP, and a long tail of banks and specialized systems. Each system served an important purpose. Together, they created significant manual work for accounting.

“We’re operating across fragmented systems, and not all of those systems integrate into our ERP,” Lindsay says. “There’s a lot of manual work behind the scenes to get that data from Workday, from Coupa, from all of these systems into our GL. It’s very manual and time-consuming.”

Every month, accountants downloaded reports, manipulated data in spreadsheets, added calculations, reformatted files for NetSuite, and passed the work to reviewers for re-performance.

For instance, The PTO accrual workflow reflected this broader pattern:

1. Download source reports

2. Manipulate and map the data

3. Add calculations

4. Reformat the output

5. Upload it to NetSuite

6. Re-perform the work during review

That sequence repeated across journal entries, entities, and systems every month. Many of these processes were also SOX controls. Every manual step added preparation time, review effort, and error risk. Each acquisition made the burden larger.

The evaluation

Zendesk was already one of the earliest enterprise adopters of AI, with a company-wide mandate from the board to operate with an AI-first mindset.

The finance organization set three priorities:

• Grow revenue without increasing headcount at the same pace

• Maintain public-company accuracy and controls

• Eliminate manual work across siloed systems

The team first evaluated the AI capabilities available in its existing finance tools. “We looked at whether there were AI capabilities in the tools we already have,” Lindsay says. “A lot of it was bolted-on AI. Those capabilities were in beta, they weren’t fully GA’d. It didn’t quite fit the bill.”

Zendesk was using BlackLine for account reconciliations and also evaluated FloQast. Both carried high costs, and the platforms felt rigid. BlackLine also required significant IT support for integrations at a time when accounting’s access to IT resources was shrinking.

The team considered building on general-purpose AI as well. “We could build our own system and put controls around it, but it’s hard,” Akshaya says. “Maintainability becomes a problem, continuity becomes a problem, and where do I put this data? I need a central system of record that is immutable. From a cost-benefit perspective, it’s better to go to a platform that does that.”

Maxima stood out because it addressed the core accounting work itself. “Maxima really hit three of the core pillars of work that we do in accounting,” Lindsay says. “Automating journal entry preparation was huge. Account reconciliations and transaction matching was the second large pillar. And the third was flux analysis and insights. That covered the core of the work that we do, especially the manual work.”

The solution

Zendesk structured its proof of concept the same way it runs its controls: deliberately, with defined use cases, measurable outcomes, and human review built into every workflow.

1. Transaction matching and reconciliations

Zendesk began with a system-to-system reconciliation that also functioned as a manual SOX control. Previously, the process relied on manual downloads and uploads into BlackLine.

With Maxima, the data flowed directly between source systems and the reconciliation workflow. The automated match rate increased from an average of 88% to more than 98%.

That improvement gave Zendesk a path to convert the manual control into an automated IT application control, or ITAC, while retaining the evidence and review structure required by internal and external audit.

The team later expanded transaction matching to two high-volume reconciliations:

• Zuora to NetSuite

• Zuora to Zendesk’s product platform

2. Journal entry preparation and posting

Zendesk selected two global use cases for journal entry automation:

• PTO accruals

• Coupa vendor accruals

Both workflows could scale across more than 25 legal entities, and both relied on source data the team had spent years making clean and reliable.

Maxima pulled the source data, applied Zendesk’s accounting logic, prepared the entries, ran validation checks, and posted approved journal entries to NetSuite. Accountants retained step-by-step visibility into how each output was created and transformed.

The vendor accrual workflow went live first for the United States, followed by a broader rollout across PTO accruals, bonus accruals, and cash journal entries for more than 20 subsidiaries.

3. Flux analysis and insights

Zendesk also moved flux analysis out of fragmented spreadsheets and into Maxima. Instead of beginning with a blank workbook, accountants now start with material variances and likely drivers already identified. The accounting team adds business context, reviews the explanations, and focuses its attention on the movements that require judgment.

Flux analysis first went live for one legal entity, followed by three additional international subsidiaries.

4. Close orchestration

Zendesk’s revenue operations team migrated close tasks and reconciliations from BlackLine to Maxima. The close orchestrator gave the team a shared system for assigning work, monitoring progress, managing dependencies, and connecting task completion to the underlying accounting outputs.

That migration alone saved approximately 30 hours per month.

Agent-prepared and accountant-reviewed

One principle governed every workflow Zendesk deployed: The agent prepares. An accountant approves.

“Before anything pushes to our ERP, it has to be approved and reviewed by a human,” Lindsay says. “The burden of proof is not on an agent. It’s on a human.” This operating model gave Zendesk the efficiency of agent-prepared accounting without weakening its control environment.

Maxima’s granular permissions also allowed Zendesk to preserve separation of duties across preparers, reviewers, and approvers.“The audit trail is very detailed and clear,” Lindsay says. “I have visibility, step by step, into how Maxima is transforming the data. And the user permissions are very granular, so we can maintain segregation of duties and fully implement our current controls environment within Maxima.”

Every transaction analyzed by an agent is logged with the source data, transformations, reasoning, and approvals behind the result. Exceptions are surfaced rather than hidden. When an agent cannot resolve a reconciliation or journal entry, it escalates the item for review instead of forcing an answer.

“From an audit perspective, we can keep the failed reconciliations out, do a manual check, and go ahead with automation for the rest,” Akshaya says. “And because we have evidence for 100% of the population, we can make the transition to full population testing. If the external auditor wants to sample, they still can, but now we have the evidence for everything.”

The impact

The results appeared in the first close cycles

• For the April close, Maxima automated 80% of PO volume in the US vendor accrual process, representing roughly one-third of the total accrual value. The workflow saved 36 hours during the most time-sensitive part of the month.

• Vendor accrual posting moved from day 4 to day minus 2, six days earlier than before.

• Instead of processing every purchase order manually, Zendesk’s accountants now focus on high-dollar and high-risk items that require professional judgment.

• Reconciliations that once depended on manual downloads now run through direct data feeds with a match rate above 98%.

• Flux analysis that once lived in Excel now begins with AI-identified drivers in one platform.

Across the workflows deployed so far, Zendesk has observed no errors. “Across what we have deployed today? Zero,” Lindsay says. “We have not seen any errors.”

The broader program is targeting:

• 100% transaction coverage

• Continuous error detection

• Automated SOX controls across the close

Zendesk’s ambition is to convert more manual controls into automated IT application controls on Maxima.

Scaling the business without scaling the team

The broader impact goes beyond individual workflows. Zendesk is building an accounting organization that can absorb continued growth, acquisitions, and new legal entities without growing headcount at the same pace. As more accounting work becomes agent-prepared, accountants spend less time assembling data and preparing entries, and more time reviewing exceptions, applying judgment, and strengthening controls.

The result is a finance organization with 2x the operational capacity, without adding headcount. For Zendesk, agentic accounting isn’t about replacing accountants. It’s about enabling a lean team to support a much larger business while maintaining public-company controls, improving accuracy, and preparing for the next stage of growth.