Skip to content

FP&A

Automate the analyst work behind FP&A.

Our AI agents investigate variances, prepare management reports, propose forecast updates, and run scenario analysis across the systems and models your finance team already uses.

Keep your ERP, planning tools, spreadsheets, and models. We do the analyst work across them.

See our agents in actionTalk to us
Illustration: an FP&A session. Asked why Q3 operating expenses are $4.2M above budget, the FP&A Agent investigates across the ERP, payroll and the FY26 operating plan: budget $42.0M, headcount +$2.1M, cloud infrastructure +$1.4M, marketing +$0.9M, professional services −$0.5M and other +$0.3M make a forecast of $46.2M, 10.0% above budget. Headcount and cloud drive 83% of it; $2.9M is recurring and $1.3M one-off. Asked whether that changes the forecast, it proposes payroll +$1.8M and cloud infrastructure +$1.1M for Q4, and no recurring change to marketing, based on the FY26 forecast model. Asked about a 60-day hiring freeze, it calculates the FY variance falling from $8.1M to $4.6M, using the forecast model and the headcount plan.

What we automate

The analyst work between your systems and the decision.

Your plans, models, actuals, and forecasts stay in the tools your team already uses. Our agents work across them to investigate, explain, update, and analyze.

An illustration of the FP&A analyst work that AI agents automate, in four connected stages: variance investigation answers what changed and why, management reporting what leadership needs to know, forecast updates what needs to change in the outlook, and scenario analysis what happens if we do X.
  1. 01Variance InvestigationWhat changed and why?
  2. 02Management ReportingWhat does leadership need to know?
  3. 03Forecast UpdatesWhat needs to change in the outlook?
  4. 04Scenario AnalysisWhat happens if we do X?

Variance investigation

Don't just see the variance. Investigate it.

Our FP&A Agent investigates material movements across your connected systems, identifies the underlying drivers, separates recurring from one-off effects, and drafts the explanation for finance review.

See the investigation
Illustration: the FP&A Agent's investigation of Q3 operating expenses, forecast $4.2M, 10.0%, above a $42.0M budget. A waterfall bridges budget to forecast through headcount +$2.1M, cloud infrastructure +$1.4M, marketing +$0.9M, professional services −$0.5M and other +$0.3M. The driver table gives each its impact next quarter and whether it recurs: $2.9M carries into Q4. Headcount is open: engineers and account executives who started ahead of budget, a merit increase six months early, and roles that were not budgeted. It was investigated using payroll journals in SAP S/4HANA, the headcount roster and the FY26 operating plan, in 5 source queries over 38 underlying records.
Q3 Operating Expense VarianceForecast vs budgetQ3 FY26All departments

+$4.2M10.0% above budget

IncreaseDecrease
$42.0MBudget
+$1.4MCloud infrastructure
+$0.9MMarketing
−$0.5MProfessional services
+$0.3MOther
$46.2MForecast
DriverVarianceNext quarter impactType
+$2.1M+$1.8MRecurring

Headcount+$2.1MQ3 FY26 · Forecast vs budget

RoleBudget startActual startBudgetForecastVariance
Software engineers (14)Engineering1 Sep1 Jul$310K$930K+$620K
Account executives (9)Sales15 Dec22 Jul$0$410K+$410K
Merit increaseAll departmentsJan 20271 Jul$0$540K+$540K
Analytics engineers (5)DataJan 20274 Aug$0$270K+$270K
Support specialists (8)Customer successNot budgeted3 Aug$0$260K+$260K
Investigated usingSAP S/4HANAPayroll journalsHeadcount rosterFY26 Operating Plan.xlsx

5 source queries38 underlying records

Cloud infrastructure+$1.4M+$1.1MRecurring
Annual conference+$0.9M$0One-off
Professional services−$0.5M−$0.3MRecurring
Other+$0.3M+$0.3MRecurring
Total+$4.2M+$2.9M

Agent analysis

Operating expenses are forecast $4.2M above budget. Headcount and cloud infrastructure account for 83% of the variance. Approximately $2.9M is recurring and will continue into Q4 unless current hiring and infrastructure commitments change.

Management reporting

The report should be the output, not the work.

Give the agent the management pack, board deck, or reporting template your team already uses. It updates the numbers, carries through the analysis, refreshes charts, and prepares commentary for review.

Your reporting format. Not ours.

See a report prepared
Illustration: the FP&A Agent preparing a finance team's own September management pack. It starts from the team's files, the Monthly Management Pack workbook as the existing template and the Q3 Board Report deck. It pulls September actuals, updates 12 KPIs, investigates 4 material variances, refreshes 6 charts and drafts the executive commentary, leaving 2 items for review. The pack, ready for review: revenue $26.5M, 2.7% above plan; gross margin 64.2%; operating expenses $15.6M, $1.6M above budget; EBITDA $1.4M; cash $41.6M. Operating expenses are open in a drawer: payroll $7.3M, cloud infrastructure $3.2M, marketing $2.1M, professional services $1.6M and other $1.4M, with cloud split into AWS $1.76M, Azure $800K, Datadog $350K and other $290K.

01Existing report

02Agent work

03Ready for review

September Management PackSeptember 2026 · Consolidated · USDReady for review
  • Revenue$26.5M+2.7% vs plan
  • Gross margin64.2%+0.6 pts vs plan
  • EBITDA$1.4M−$1.0M vs plan
  • Cash$41.6M+$1.1M on August

Select operating expenses to drill down

Executive commentary

Revenue finished 2.7% above plan, while operating expenses remained above budget. Higher personnel and infrastructure costs account for most of the variance.

Operating expensesSeptember 2026$15.6M
  • Payroll$7.3M
    • AWS$1.76M55%
    • Azure$800K25%
    • Datadog$350K11%
    • Other$290K9%
  • Marketing$2.1M
  • Professional services$1.6M
  • Other$1.4M
View source records

Forecast updates

Keep the forecast current without rebuilding it.

The agent identifies what has changed since the last forecast, investigates the financial impact, and proposes updates to the model your team already uses. Finance reviews the proposed changes before accepting them.

Your planning software
Stores and calculates the forecast.
Our FP&A Agent
Figures out what should change and why.
See a forecast update
Illustration: the FP&A Agent working in a finance team's own forecast model, FY26 Forecast Model.xlsx, version 12, last updated 1 September. Since then it has found three changes: hiring that ran ahead of the headcount plan, +$1.8M in Q4, from payroll and the headcount plan; committed cloud infrastructure spend that increased, +$1.1M in Q4, from the contract and the ERP; and the annual conference, unbudgeted Q3 spend that is one-off, $0 in Q4. It proposes the Q4 updates: payroll from $22.6M to $24.4M, cloud infrastructure from $8.5M to $9.6M, and marketing unchanged at $6.3M, a recurring Q4 adjustment of +$2.9M, for finance to review. A chart shows FY26 operating expenses by month, actual to September and budget for the year, with the current forecast and the forecast with the proposed updates for Q4, ending December at $16.6M.

Scenario analysis

Ask the question. The agent does the model work.

It gathers the relevant plan, assumptions, commitments, and cash position, then runs the question through defined financial logic or the model your team already trusts.

See a scenario run
Illustration: a hiring scenario. Asked whether the company can hire 40 more engineers from January, the FP&A Agent runs the question through the team's FY26 forecast model, version 12, its headcount plan and its cash plan. The assumptions: an average loaded cost of $145K from the headcount plan, a start on 1 January, a four-month hiring ramp and $12K of recruiting cost per hire, against a minimum operating cash of $20M from treasury policy. Each option is tested against three limits: no more than $0.5M added to the FY26 forecast, no more than $3.0M added to the FY27 plan, and cash of at least $20M. The base case and 20 hires pass all three. 40 hires pass FY26 and cash, at a low of $20.7M in May 2027, but add $4.8M to FY27, $1.8M more than the plan allows. So 40 hires do not fit the plan: they stay within this year's forecast and cash, but exceed the FY27 plan. 60 hires fail all three.
Engineering hiring scenarioFY26–FY27ConsolidatedUSD

Can we hire 40 additional engineers starting in January?

Models usedFY26 Forecast Model.xlsx v12Headcount Plan.xlsxCash Plan.xlsx

Scenario assumptions

  • Additional hires
  • Start date1 Jan
  • Average loaded cost$145KFrom Headcount Plan
  • Hiring ramp4 months
  • Recruiting cost / hire$12K

Limit

Minimum operating cash$20MTreasury policy
ScenarioFY26 forecastFY27 planCashOutcome
Up to +$0.5MUp to +$3.0MAt least $20M
Base casePass+$0Pass+$0Pass$22.6MWithin plan
+20 hiresPass+$0.24MPass+$2.4MPass$21.6MWithin plan
+40 hiresPass+$0.48MFail+$4.8MPass$20.7MFY27 exceeds plan
+60 hiresFail+$0.72MFail+$7.3MFail$19.7MExceeds limits

Agent summary

The 40-hire scenario stays within the FY26 forecast and minimum cash, at $20.7M in May 2027, but adds $4.8M to FY27 operating expenses, $1.8M more than the plan allows.

Calculated from defined assumptions. Not generated by AI.

Trustworthy by design

AI explains it. Finance math calculates it.

We use AI to understand questions, gather context, investigate drivers, and communicate the answer. Financial calculations run against defined models, formulas, assumptions, and source data.

A diagram of the product's two layers. The AI layer takes a question, understands its intent, finds the relevant data, investigates the drivers, explains the result and generates the commentary. It hands calculations to the models and calculations layer, which takes the existing model, budget, actuals, contracts, assumptions and cash through a defined calculation, made of spreadsheet formulas, SQL logic, business rules, decimal arithmetic and versioned assumptions, and returns the budget, forecast, cash and scenario results the explanation uses.

The explanation can be AI-generated.The numbers are calculated.

Data lineage

Every number has a source.

Move from the executive summary to the transaction, contract, spreadsheet cell, or system record behind the number.

Illustration: September operating expenses of $15.6M traced to a record. Cloud infrastructure is $3.2M of them: AWS $1.76M, Azure $800K, Datadog $350K and other $290K. Behind AWS is AWS-092826, one of nine AWS invoices for September, for $218,420, read from SAP S/4HANA, with its source, last sync, calculation, the reports it is used in and its owner.

Ask follow-ups

Keep asking.

Each answer is a number you can question, not a paragraph to read.

Illustration: a CFO's conversation with the FP&A Agent. How are we tracking against budget? $4.2M above plan. Why? Headcount and cloud account for 83%. How much is recurring? $2.9M. Does that change our forecast? Three proposed updates, adding $2.9M to Q4: payroll +$1.8M, cloud infrastructure +$1.1M, marketing unchanged. What if we freeze hiring for 60 days? The FY variance falls from $8.1M to $4.6M. Add that to the board report. Drafted in the Q3 Board Report, slide 7, ready for review.

Integrations

Work across the FP&A stack you already have.

Connect the systems that hold actuals, plans, headcount, pipeline, cash, and supporting detail. Our agents work across them without requiring finance to rebuild its stack.

A diagram of what the FP&A Agent works across, by purpose: ERP (SAP S/4HANA, NetSuite, Business Central), data warehouse (Snowflake, PostgreSQL, Oracle Database), CRM (Salesforce, Zoho CRM), banking and spend (Plaid, Ramp, Brex, BILL), plans and models (budgets, forecast models, spreadsheet plans and warehouse tables, as files or tables), and HR and payroll (payroll journals, HR exports and headcount plans, from the ERP or as exports).
View all integrations

Control

Built for decisions finance can defend.

Keep sources traceable, assumptions visible, calculations reproducible, access controlled, and humans in control of material decisions.

  • Role-based access

    Who sees which reports and models is controlled by role.

  • Source permissions

    Agents read only what each connection allows.

  • Human approvals

    Material changes wait for a named reviewer.

  • Version history

    Every revision remains tied to the run that produced it.

  • Calculation history

    Each result records inputs, logic, and version.

  • Audit trail

    Every question, query, proposal, and approved change appears in the activity log.

Read the security page
Illustration: a hiring scenario under review, 40 more engineers from January. It passes the FY26 forecast and cash limits and fails the FY27 plan. Its analysis details: five data sources connected, five assumptions defined, calculation version 3.2, prepared by the FP&A Agent, reviewed by Sarah Chen, the analysis verified, last updated 30 September 2026.

Case studyKKR-backed European education group

A KKR-backed education group connects management reporting and finance operations.

More than 120 management questions mapped to the systems that answer them, across four ERP instances, a CRM and academic records, and an invoice processing module developed and awaiting testing.

“Empty Spaces has become our new operating system. It gives us a clearer view of our operations and surfaces the insights we need to run the company better.”
Chief Strategy Officer

Questions

The questions we get asked first.

  • What is FP&A by Empty Spaces?

    Our AI agents automate analyst work across FP&A, including variance investigation, management report preparation, forecast updates, and scenario analysis. They work with the finance systems, plans, spreadsheets, and models your team already uses.

  • What do you automate that our planning software doesn't?

    Planning software stores and calculates budgets, plans, forecasts, and models. Our agents automate the analyst work around them, such as investigating why results changed, gathering context across systems, preparing reporting, identifying forecast updates, and running financial questions through existing models.

  • Do you replace our budgeting and planning software?

    No. Budgets, forecasts, and models can remain in the systems your team already trusts. Our agents work across them and automate the analysis and manual work around them.

  • How does your FP&A Agent investigate a variance?

    It compares the financial result with the relevant budget, forecast, or prior period, identifies material drivers, investigates supporting data across connected systems, separates recurring from one-off effects, and prepares an explanation for finance review.

  • Do you replace our ERP?

    No. We work across the ERP and other systems finance already uses.

  • Can your agent prepare our existing management pack or board deck?

    Yes. It can work with existing reporting structures and templates, update the underlying figures, incorporate current analysis, refresh supporting content, and prepare the output for review.

  • Can your agent update our existing forecast?

    It can identify business changes that affect the forecast, calculate their impact using defined financial logic, and propose changes for finance review.

  • Do you generate forecasts with AI?

    No. AI can investigate changes, gather context, and explain results. Financial outputs come from defined formulas, models, assumptions, and source data.

  • Can your agent run scenario analysis?

    Yes. It can gather the required assumptions and financial context, run scenarios using defined models or calculation logic, and explain the resulting impact on budget, forecast, cash, and other financial constraints.

  • Can your agent work with our Excel models?

    Yes. Existing spreadsheet models can remain part of the FP&A workflow; we don't require them to be rebuilt.

  • Can every number be traced back to its source?

    Yes. Outputs can remain connected to supporting transactions, records, contracts, spreadsheets, and source systems.

  • What requires human approval?

    Teams can define which outputs the agent prepares automatically and which changes, forecast updates, or other material actions require review before they are accepted.

  • Who is your FP&A Agent for?

    It is designed for CFOs, finance leaders, FP&A teams, controllers, finance managers, and analysts responsible for reporting, analysis, forecasting, and financial decision support.

Automate the work between the numbers and the decision.

Let our agents investigate variances, prepare management reports, propose forecast updates, and run scenario analysis across the finance stack your team already uses.

See our agents in actionTalk to us