The flagship conference ‘Finance. Taxes. Investments’, a major annual business event that brings together professionals in law, tax, audit, and consulting, was hosted by FBK on 16 October.
Here are key insights from the joint presentation by Elena Tyutyunnikova, Senior Partner at FBK, and Dmitry Livshin, General Director of CYBER Business Consulting.
The core paradox is that 80% of employees report increased personal productivity thanks to AI, yet 96% of executives see no significant impact on business metrics. 95% of AI experiments have no measurable financial effect. Why does this happen?
Three gaps between expectations and reality
1. Point solutions instead of function transformation.
Most companies implement AI as a set of isolated solutions layered over existing processes. Leaders, however, act differently: almost three-quarters of them adapt their processes to AI.
2. AI on top of poor data architecture.
AI does not fix fragmented systems or duplicate master data—it makes them more visible. Leaders invest in data and knowledge bases long before implementing the technology.
3. Workforce deskilling.
50% of top managers notice signs of skills loss among employees. The question arises: how can a professional gain experience if an agent handles the routine tasks?
Where AI agents give results first
- Accounts payable – allocating invoices, approving payments below the threshold.
- Reconciliations – collecting transactions from all systems, identifying the causes of discrepancies.
- Period close – sequencing tasks.
Elena Tyutyunnikova shared four practical use cases of an AI agent within the accounting function, where the agent operates directly within the company’s 1C environment. The expert here defines the methodology once, and the agent replicates it for any given period.
Case 1. Income tax control.
Before the period close, the agent performs control procedures: it checks current and deferred tax, and temporary differences. The output is a list of discrepancies with amounts and source references.
Case 2. Express analysis of accounting records.
The agent scans the transaction pool for anomalies that deviate from the company's normal operating model: new and rare account correlations, large and manual entries, adjustments to closed periods, negative balances. Instead of a standard trial balance, the accountant receives a short list for review. The decision on whether an error exists is made by a human.
Case 3. Financial reporting review.
Following the methodologist's rules, the agent reconciles the balance sheet, statement of income and changes in equity, and notes against each other and against the system data. What was previously checked in Excel is now verified directly against the accounting data.
Case 4. Amended tax returns.
The agent monitors changes in closed tax periods. It analyses the documents, determines whether an amended return is needed, calculates the impact, and notifies the accountant.
6 steps that leaders do:
- Look at the function as 5–7 end-to-end processes, rather than as separate departments.
- For each process, apply three levers in sequence: first change the process, then automate it, and only then connect AI.
- Decide what to do within the system, what to assign to a robot, and what to assign to an agent.
- Invest in data and the knowledge base – this is the context without which AI cannot work.
- Focus on developing and retaining employee skills.
- Before launch, define the target outcome and a plan for making use of the freed-up time.
FBK provides a full spectrum of advisory: from transforming finance function and implementing the AI agent to automating accounting. Your projects can be discussed in detail during a consultation.
