Embracing Agentic AI in Finance
The discourse around Artificial Intelligence continues to evolve at a breathtaking pace, with Agentic AI emerging as the next frontier in business transformation. For CFOs, this offers both unprecedented opportunities and significant challenges in harnessing this advanced technology. As Chief Financial Officers navigate the complexities of digital transformation, understanding and strategically adopting Agentic AI is no longer optional; it’s an imperative to maintain relevance and drive tangible business value.
While many organizations are still grappling with integrating Generative AI (Gen AI), Agentic AI represents a significant leap forward. Gen AI excels at generating new and unseen content based on its training data. For instance, you could ask it to “analyse last quarter’s sales” and it would likely do a relatively great job, providing analysis, tables, graphs, and even recommendations.
Beyond Generative AI: Understanding Agency
However, its capabilities are restricted when faced with multi-step, interconnected problems. If you were to ask it to “analyse sales and identify root-casue” the output would be highly restricted to its trained data and most likely incorrect, as it typically ignores broader organizational context and input from other functions like sales or marketing. This is more like a co-pilot fucntionality.
However, as we move from co-pilot to full action-oriented AI, this same complex task by an agentic AI system will be done in a significantly independent and interconnected manner, much like human teams. Imagine one agent downloading actual results from an ERP system, while simultaneously others gather inputs from the sales team on market conditions and drivers impacting sales and the marketing team on innovation launches. Another agent would then collate all this diverse information into a single coherent view. Humans remain in the loop during these agentic workflows only at critical points or for validations.
This autonomous orchestration of different agents, working collaboratively, is what makes this technology immensely powerful. The rapid progress is significantly fueled by Large Language Models (LLMs) becoming increasingly sophisticated in reasoning and calculations, allowing them to handle complex tasks with fewer human touchpoints. Furthermore, these models are no longer restricted to text inputs; they can work remarkably well with images, audio, and videos, enabling multiple AI systems to work together to solve complex, multi-step business problems.
For more expert lessons on Agentic AI for finance leaders dig into Anna Tiomina’s CFO Talk: Revolutionizing CFO Operations with AI Agents
Despite being in its infancy, Agentic AI is already demonstrating substantial efficiencies across various industries. We have seen firms like Alaska Airlines, Puma, Wiley, and Best Buy using AI agents to enhance customer experiences and redesign operations. Although debates persist regarding the true level of autonomy in every application, early deployments offer compelling evidence of its potential to create a tangible business impact. These advancements underscore how powerful it is when different AI capabilities are converged to tackle complex workflows.
Where do we stand?
However, it’s crucial to recognise that Agentic AI may not be suitable for every business problem. Problems that are less complex and straightforward might still benefit more from traditional AI approaches or even low/no-code automation. The ROI might not justify the use of AI agents for every multi-step problem, and having a human in the loop might be a more feasible option.
While Agentic AI promises transformative potential, many businesses are still struggling to realise significant impact from current Gen AI implementations, with 80% of AI projects reportedly falling short of objectives. This raises valid questions about the practicality of deploying Agentic AI when fundamental issues with existing AI models remain unresolved. Furthermore, finance functions often grapple with years of accumulated complexity in processes, disparate data, and legacy technology stacks.
The CFO’s Path Forward
However, inaction is not an option. CFOs must strategically embrace this evolution to prevent the finance function from becoming irrelevant. The journey to embedding Agentic AI requires a structured approach:
- Build a Strategic Roadmap: Avoid jumping into isolated solutions or collecting use cases. Instead, establish a clear digital transformation strategy with a compelling vision that articulates why this transformation is necessary and how Agentic AI fits into the future state of finance. This involves creating a clear view of both the current and future states to build a roadmap.
- Fix the Data Foundation: Data quality remains the biggest hurdle to AI adoption, cited by 85% in KPMG’s AI Quarterly Pulse Survey 2025. A strong data foundation, including cloud migration, retiring legacy systems, and robust data pipelines, is crucial. This foundational work, though seemingly slow initially, leads to “hockey-stick growth” in business value as reusable datasets become available, enabling faster scaling of digitalisation efforts. CFOs must prioritise co-creating a data strategy with C-suite peers, focusing on data foundations, democratisation, and governance frameworks.
- Identify Momentum Builders: Experimentation is key, but it should be thoughtful and controlled. Start with the least risky processes where potential errors won’t cause massive disruption. For example, delegate workflow coordination in financial planning or closing processes to AI agents, while humans remain in the loop for critical reviews and approvals of financial entries. This “test and learn” approach builds confidence before a broader rollout, by delegating less risky collaboration pieces to AI agents while tasks impacting financials remain under the human domain.
- Prioritize Process Excellence: Automating an inherently complex or flawed process will only amplify its complexity and can potentially expose the organisation to unknown risks. Identify high-complexity processes, conduct redesign sprints (internally or with external help), and categorize improvement opportunities across people, process, system, and data workstreams, integrating them into your digital transformation roadmap. The quality of Agentic AI directly depends on the standardization and complexity of the underlying processes it supports.
- Collaborate Externally and Build Internal Capability: The AI landscape is ambiguous, complex, and rapidly changing; no organisation can navigate it alone. Engage with external vendors, join industry networks, attend tech conferences, and encourage C-suite visits to learn from early adopters and understand risks. While well-reputed external consultants can assist, the ultimate goal must be to build robust in-house capabilities rather than solely relying on external support for implementation.
** This article includes insights from Tariq’s book, “Reimagine Finance: The CFO’s Leadership Playbook for the Age of AI, Data, and Digital” (Wiley), which is coming out on Sep 9th, 2025. Click to Pre-order. **
A Fundamental Redesign
Ultimately, embedding Agentic AI is not merely about retrofitting new technology into existing workflows. It demands a fundamental redesign of current operating models and ways of working within finance. This presents CFOs with a unique opportunity to lead this revolution, balancing ambitious goals with realistic returns and avoiding the pitfalls of technology hype.
In a related CFO Talk, Tariq sheds a bright light on how process mining is driving big improvements in CFO operations in Process Mining - Revolutionizing Corporate Finance with Tariq Munir
Identify your path to CFO success by taking our CFO Readiness Assessmentᵀᴹ.
Become a Member today and get 30% off on-demand courses and tools!
For the most up to date and relevant accounting, finance, treasury and leadership headlines all in one place subscribe to The Balanced Digest.
Follow us on Linkedin!
