Why Spreadsheet-Based Tax Provision Breaks at Scale (And What to Do Instead)

Key takeaways

• Spreadsheet-based provision processes can produce accurate calculations, but they fall short on consolidation, standardization, and knowledge continuity as organizations add entities and jurisdictions.
• AI adds the most value in provision when applied to review, documentation, and monitoring rather than the underlying rules-based calculation.
• A connected, controls-oriented system reduces manual work, strengthens audit readiness, and builds the data foundation organizations need before layering in AI.

As organizations add entities and enter new jurisdictions, many tax departments continue to assemble their provision from a patchwork of spreadsheets, disconnected tools, and manual workarounds. These processes may produce the right number, but they rarely produce a reliable path from data to disclosure.

That weakness was the focus of a recent Bloomberg Tax webinar, Creating a Scalable Provision Workflow from Data to Disclosure.

The discussion brought together Nick Frank, founder of TAXPlanations, Darian Harnish, a partner at RSM US LLP, along with Mackda Dinberu, senior manager of solutions consultants at Bloomberg Tax, and Holly Hanley, senior solutions consultant for Bloomberg Tax.

In this article, we share some of the highlights from that conversation to unpack where legacy provision processes break down and what tax leaders should prioritize as they modernize.

Where spreadsheet-based provision breaks down

In Nick Frank’s view, the cracks typically appear as organizations grow through acquisitions and new jurisdictions and try to replicate what already works.

A single spreadsheet built to calculate a book-to-tax difference gets copied for the next entity, and the next, until the organization is managing dozens or even hundreds of individual workbooks.

Tax provision processes must scale with business growth

“Each entity has its own workbook,” Frank said, “and you have to come up with some way to consolidate all the data.”

Because each spreadsheet produces the correct answer in isolation, tax teams are often reluctant to change the underlying process, even as the burden of pulling everything together grows.

As the business keeps expanding through new contracts, entities, and jurisdictions, a provision process needs to absorb that growth without depending on the assumption that next year will be simpler.

“Scalable to me means you can absorb those into your process in a relatively seamless way,” Frank said, rather than treating each year’s complexity as a one-off exception.

Hear Nick Frank, founder of TAXPlanations, explain how adding entities and jurisdictions can turn spreadsheet-based provision into a growing consolidation and standardization challenge. [2:33]

Spreadsheet-based processes create key-person risk

Darian Harnish pointed to a related risk from depending on institutional knowledge. Ad hoc, spreadsheet-based processes tend to rely heavily on a single, long-tenured employee who understands the specific formulas, hard-coded adjustments, and roll-forward logic embedded in the model.

“So much depends on one person who’s been running that process for a long time,” Harnish said.

When that person leaves, retires, or moves on, the incoming team is left reverse-engineering a process that was never documented, potentially delaying reporting.

“You think these processes work up until you don’t,” Harnish said. “And then it’s several nights of 4 a.m. in a row. And that’s not the time to be making sure that your footnote ties out.”

Signs your provision process depends too heavily on institutional knowledge:

• Critical formulas are understood by only one person
• Adjustments are hard-coded rather than documented
• Roll-forward logic varies by workbook
• New team members must reverse-engineer prior-year files
• Reporting could be disrupted by employee turnover

See how tax provision software connects calculations, research, and AI-enabled intelligence to reduce manual work and support audit readiness.

Data quality can create downstream tax provision problems

Provision workflows are only as reliable as the data feeding them, and the data problems compound as organizations scale.

Harnish traced many of these issues back to the enterprise resource planning (ERP) implementation stage, when decisions about legal entity structuring and data capture don’t necessarily consider tax provision concerns.

Organizations that operate as a single integrated segment for accounting purposes must often unwind that structure later. For example, divesting a legal entity or jurisdiction-specific reporting requirements, such as state net operating loss disclosures, requires a level of detail the provision model can’t support.

“That’s exactly where it starts to break down,” Harnish said, adding that companies that avoid shortcuts during implementation can pull clean disclosure data directly from the provision rather than reconstructing it from return carryforward schedules with deadlines looming.

Standardized tax data makes consolidation easier

Frank identified another recurring data issue stemming from inconsistent naming and descriptions across entities and systems. A deferred tax item might be labeled “fixed assets” in one entity’s provision and “depreciation” in another.

An entity might have a different name or alias across the ERP, compliance software, provision system, and internal shorthand. Reconciling those inconsistencies at consolidation is its own governance challenge.

“Having good governance around how you’re going to move things through that process” matters because early structuring decisions “manifest themselves in the footnote at the end of the day,” Frank said.

Better tax data governance reduces downstream rework

Harnish added that seemingly minor implementation shortcuts, such as how much state-level detail you capture in the provision model, create outsized downstream costs, adding hours to the return-to-provision process later.

“Clean data that matches up clearly to the return is a gold standard,” Harnish said. “It might take a few extra hours upfront to implement it, but it pays huge dividends on the backend.”

These data challenges aren’t incidental. With ASU 2023-09’s expanded rate reconciliation disclosure requirements, evolving state conformity and decoupling positions, and the data demands introduced by Section 174 capitalization rules under the One Big Beautiful Bill Act (OBBBA), the volume of data that must flow correctly through the provision to the footnote has only increased.

Bring more control to every stage of the provision process. Discover how Bloomberg Tax Provision helps tax teams standardize processes, reduce reconciliation, and stay connected to the data and intelligence behind every number.

Where AI fits in the tax provision process – and where it doesn’t

Historically, tax departments have hesitated to adopt tools that handle every use case. AI, by contrast, has seen rapid adoption despite inherent unpredictability.

Tax provision calculations require consistent, rules-based logic

“It’s not great at doing anything in particular in the sense that it’s amazing, yet I can’t be assured if I put this in on one side, I get that out on the other side.”

Frank is still not confident in using AI for tax calculations, where a given input reliably produces a known output.

“[Whereas] with provision software, I can tell you if you put in a number what’s going to happen on the other side,” Frank said.

He recommends using AI for what it does well and shore up its limitations with process and human review, rather than expecting it to replace deterministic calculation logic.

Harnish reinforced that distinction. Tax calculations, and the FASB ASC 740 codification underlying them, are fundamentally rules-based.

“Tax is generally a rules-based logic,” Harnish said. “You need the same output every time.”

Hear Darien Harnish, partner at RSM, explain why AI is better suited to provision review, documentation, and monitoring than replacing consistent, rules-based calculations. [2:55]

AI can strengthen provision review, documentation, and monitoring

AI is a better fit for day-to-day provision, flagging workpapers that may be outdated following a tax law change, and drafting standard process narratives to support internal controls and SOX compliance.

Bloomberg Tax is applying AI in this way within its connected tax workflow. In Bloomberg Tax Workpapers, the developments tracker can scan a workpaper to surface recent topics that may be relevant to the calculations at hand, helping tax professionals spend less time manually searching for developments.

Users can then ask follow-up questions through the AI Assistant, which draws on trusted Bloomberg Tax Research to provide additional context and help them quickly assess what may require attention.

The result is a more practical role for AI in provision work: not replacing the rules-based calculations tax teams rely on, but bringing relevant intelligence closer to the work so professionals can identify changes faster, investigate them efficiently, and apply their judgment where it matters most.

Tax AI requires strong data security and standardization

Data security and privacy must come first, Harnish noted, given that provision work involves sensitive financial statement data.

Organizations evaluating AI-enabled tax tools should understand how the vendor handles and protects data before adoption.