Tax strategy and planning: what to automate, what stays human, and the data it depends on

Software can now maintain tax master data, pull figures from the ledger, apply rates and run scenario models across jurisdictions. Choosing a strategy, weighing risk and signing off positions still need a tax professional. None of the automation holds up unless legal entity, ownership and account mapping data is clean.

Maintaining tax master data

This is the most mechanical part of the process and the best place to start. Master data here means the legal entity register, ownership percentages, tax registrations, jurisdiction codes, filing calendars and the mapping from the chart of accounts to tax categories. Rule-based tools handle much of it well. They can flag a new entity in the ERP that has no tax profile. They can track rate changes published by authorities and push them into the calculation engine. They can also check that every account maps to a tax treatment.

AI adds value at the edges. It can read incoming legislative updates and suggest which entities might be affected. A person still has to confirm the suggestion. A wrong rate in master data quietly corrupts every downstream number, so approval of changes should stay with a named owner.

Setting strategy for each jurisdiction

Foreign, national, state and local strategy is where judgement dominates. Software can gather the inputs. It can summarise where profits are booked, where people sit, where intellectual property is held and which incentives exist. It can draft comparisons of treaty positions or credit eligibility.

The decisions themselves belong to people. How aggressive a position to take, whether a structure survives scrutiny on substance, how a choice will look to an authority or to the public: these depend on the organisation's risk appetite and its history with each tax office. Generative tools can misstate law with confidence. Treat their output as a first draft that a qualified reviewer checks against primary sources.

Consolidating and optimising the total plan

Once each jurisdiction has a view, the plan has to be pulled together into a single effective tax picture. Here automation earns its keep again. Consolidation engines can roll up entity forecasts, eliminate intercompany flows and recalculate the group rate under different scenarios. Modelling a change to transfer pricing, a restructuring or a new market entry used to take weeks of spreadsheet work. A well-built model now reruns it on demand.

Optimisation is a different matter. A model will happily recommend the scenario with the lowest tax. It will not know that the operating business cannot move those people, or that the board has ruled out a particular structure. Someone has to translate the numbers into a recommendation that fits commercial reality and then defend it.

What the data has to look like first

Most failed tax automation projects trace back to data, not to the tool. Before switching anything on, check these conditions:

• A single legal entity register that finance, legal and tax all accept as true, with effective dates for changes in ownership.
• Account mapping at a level tax actually needs. A general ledger built for management reporting often lumps together items that are taxed differently.
• Intercompany transactions tagged consistently on both sides, so eliminations and transfer pricing analysis reconcile.
• Forecasts by legal entity, not only by business unit or segment.
• A clear owner for each data element and a record of who changed what.

If forecasts only exist by segment, the plan will rest on allocations someone made up. Fix that upstream before buying a modelling tool.

Questions to ask the people who run it

Documented procedures for tax planning tend to describe an orderly cycle. The real work is usually messier. These questions surface the gap:

• Where does the team actually get entity profit forecasts, and how often are they adjusted by hand before use?
• Which spreadsheets sit outside the official systems, and who would be stuck if they disappeared?
• When a new entity is created, how does tax find out? Has an entity ever been missed?
• Which account mappings does the team distrust, and what do they do to work around them?
• How are rate and law changes tracked today? Is it one person's inbox?
• Who has authority to approve a planning position, and is that written down anywhere?
• What questions from auditors or authorities have been hardest to answer from the current records?
• Which parts of the cycle get rushed at year end, and why?

The answers usually show that master data upkeep is informal and that the consolidation model depends on a single expert. Those are the places to automate first and to document before anyone leaves.

Where to draw the line

A workable split puts software in charge of collecting, checking and calculating, while people choose, approve and explain. Keep humans on any output that commits the organisation to a position with an authority. Keep an audit trail on every automated change to master data. Review the split whenever the group structure changes, because a new acquisition can break assumptions that the tools were built on.

APQC's Process Classification Framework® (PCF) is an open standard developed by APQC, a nonprofit that promotes benchmarking and best practices worldwide. To download the full PCF or to view definitions and measures, please visit www.apqc.org/pcf.