Q&A with Abdullah Almsaeed, Co-founder & CTO of Financial Cents

Abdullah Almsaeed, co-founder and CTO of Financial Cents, explains where AI is already useful for small accounting and bookkeeping firms, why firm-specific context and auditability matter, and how owners can begin with low-stakes tasks.

AI in the Accounting Industry

When you think about where AI is truly helping small accounting and bookkeeping firms today (versus where it’s still more hype than help), how would you describe that line?

I believe that AI can give firms superpowers. It is an amazing tool to add to your arsenal. I’ve seen firms use AI successfully in streamlining client communication, organizing documents, automating data extraction, and even preparing drafts of complex tax returns. AI becomes “hype” when there is no easy and fast way to validate its work. Because AI can and will make mistakes, it’s the firm’s responsibility to put systems in place that evaluate its work. For example, apps that advertise “autonomous close” but don’t offer error detection and review tools will most certainly not be useful.

What’s something about building AI into an accounting-focused software that’s harder than people might realize?

What is considered “correct” in one firm isn’t necessarily the same for another. For instance, there is no clear answer to, “Which account does this transaction go into?” There’s only, “The way this firm does it for this client.” This makes building a successful AI product really hard, but this is also exactly where AI can shine. It’s a problem that we can finally solve, because AI can be trained and fine-tuned for each client.

AI in Financial Cents

Financial Cents builds AI specifically for small and growing firms. How does this differ from building it for a large enterprise?

The biggest difference is that an enterprise firm has the resources to dedicate toward learning, configuring, and setting up the platform. Someone on the team can champion its adoption, rollout plan, training sessions, etc. For a smaller firm, you get one chance to impress the owner who is likely busy and trying it for the first time between a client call and a payroll deadline. If it doesn’t work as expected, it’s mentally filed under “AI stuff that wastes my time.” That fundamentally changes how we approach AI in the product. We spend a lot of time designing features with a clear value statement, easy to use and fast to implement. We call it “AI that just works.”

There’s enormous pressure right now to “add AI” to everything. How do you decide what actually belongs in the product versus what’s just noise?

We’re experimenting with AI in every aspect of our app. It only moves to development if the value is clearly stated. Nobody just wants AI. They want to solve a problem. So we always ask if this new feature removes a step from something they do daily, or if it adds another place they have to go. It also depends on whether we can design it to be easy to adopt with minimal to no setup.

What has been the response to the AI agents in Financial Cents? Any metrics you can share?

The response has been amazing! We get people going out of their way to email us about how much they’ve been enjoying it and how useful they’ve found it to be. Since the initial release of our AI agents in July, 95% of firms who experienced an agent kept using it every day since. We’ve so far processed over half a million documents with our document naming and validation agents. If you think about that for a minute, it means we’ve saved a collective 25,000 human hours (assuming each document needs three minutes to review, rename, and file in the right folder). This is the kind of impact we’re looking to achieve, and it’s only the beginning.

Accountants worry a lot about trust (i.e., whether the AI will get it right, and who is accountable when it fails). How does that shape what you’re willing to automate in the product versus what you believe should remain solely in the user’s hands?

That is a very valid concern. This is exactly why systems of record like Financial Cents are important. It’s a place where you can track and audit the behavior of AI agents. Every agent we build ships with audit logs and settings that allow users to pick what they want to completely automate versus when they want to keep a human in the loop. We also design each agent’s actions to be easy to evaluate and correct in case of mistakes.

How the Financial Cents Team Uses AI

How is your own team using AI day-to-day? Which tools and use cases have and haven’t stuck?

In the product and engineering department, AI touches almost everything at this point. Starting from the usual call recordings and email summaries to more complex tasks like designing interfaces, writing code and setting up acceptance criteria. We use Claude.ai for almost all of it and Fellow.ai for call recordings.

AI is changing fast. How do you and your team stay current without burning out on it?

Engineers are used to software evolving fast and new technologies emerging every day, so I feel like it’s less of a shock to us. However, what really helped us was that we committed to spending at least three months learning any given tool before moving on to something new and shiny. That allowed us to gain a deeper understanding of our tools and helped us get to the point where we’re realizing their true value.

What are some examples of things your team has built or accomplished with AI that you are most proud of?

One clear success metric is that our overall output has increased by over 1,000%. Yes, you read that number correctly. It’s an insane increase in productivity that allows us to focus on more complex problems and solve the most pressing issues for our clients fast. To get there, we’ve spent a lot of time developing our own internal tools and workflows to help us harness the power of AI. How we plan, design, and develop today is completely different than a year ago.

On a personal note, what’s one AI tool or habit you’ve picked up that you’d really miss if it disappeared tomorrow?

Might be obvious at this point, but I can’t live without Claude anymore.

AI in the Future

If a firm owner reading this wanted to start experimenting with AI but felt intimidated or overwhelmed, what’s the first small thing you’d tell them to try? What would you tell them to ignore for now?

The fact that AI has virtually no limits makes it hard to approach or know where to start. I started by asking Claude to do very complex tasks that I am very familiar with and honestly very opinionated about. I hated its output and wanted to move on. I realize now that’s the wrong approach. You should most certainly start with tasks that you care less about and ask Claude to automate them. For example, “Archive all advertising emails from my Gmail account.” Or, “Organize these documents into folders by document type.” By starting small, you’re building the skills needed to know how to prompt the AI well and developing the trust needed for you to want to automate more. Soon, you’ll be building AI powered workflows that make you feel superhuman. To me, it’s an exciting technology that will have a huge impact on everything, so it’s really important for us to understand it and harness its potential.