Where It Started

As a teenager, he looked at Google and Facebook and wanted to build something too.

He learned to code. He built a blogging platform where people could write and earn money.

He was young, had no experience in distribution, and the platform died. But the direction was set from that moment.

He went on to work as a backend engineer at Flutterwave, one of Africa’s most significant fintech companies, where he engineered the checkout systems used by millions of people.

That is not a generic engineering role. Checkout infrastructure sits at the point where payment intent converts to transaction completion.

The reliability requirements are extreme. The failure cost is immediate and measurable. The scale is unforgiving.

After four years at Flutterwave, he consulted for AI companies. Then he noticed something that had been in front of him for years.

The Observation That Became the Company

At Flutterwave, he worked alongside many people who dealt with data in spreadsheets. He would watch them pull data manually. This was 2019 and 2020, before AI had matured enough to change the picture.

Years later, as large language models improved and started demonstrating genuine capability with structured data, he saw the same problem differently.

The people he had watched doing manual spreadsheet work were still doing manual spreadsheet work. The tools had not reached them in any meaningful way.

He also noticed a gap in how AI companies were positioning their products. Everyone was building general-purpose tools.

Nobody was building specifically for spreadsheets, even though Excel alone has over one billion users, making it one of the largest and most consistent professional software audiences in the world.

That combination, a massive, underserved user base, a technical capability that had just matured enough to serve it, and a competitive landscape distracted by generality, was the founding observation behind Decide.

The Core Problem

The problem Decide is solving is not that spreadsheets are hard to use. Millions of professionals use them every day with reasonable competency.

The problem is that the full range of spreadsheet work, cleaning messy data, building financial models, extracting tables from PDFs, creating dashboards, running web scrapes into structured formats, constructing cap tables, building payroll models, is time-consuming, repetitive, and cognitively demanding in ways that have nothing to do with the strategic thinking the professional should be spending their time on.

A KPMG analyst who needs to reconcile three months of transaction data across multiple formats is not doing high-value work when she spends four hours cleaning the dataset before analysis can begin.

An investment professional who needs a cap table built from a term sheet is not adding value by doing the mechanical construction work himself.

Decide is built for the gap between what these professionals need to produce and how long the mechanical work to get there currently takes them.

The accuracy benchmark matters in this context specifically.

An AI tool that produces wrong outputs in financial modelling or data analysis is not just unhelpful. It is dangerous. The 82.5% accuracy on Spreadsheet Bench, described by the founder as making Decide one of the most accurate in the world for this task, is not a vanity metric.

It is the minimum threshold of trust that makes the tool usable for professionals whose work carries real consequences.

The Strategic Decision Layer

One of the most instructive strategic moment in this company’s history is not the founding. It is the naming.

When Decide launched, the positioning was data analysis. That framing was accurate but too broad.

Data is an enormous space. Data analysis tools are everywhere. Potential users encountered the description and did not immediately understand what they were being offered or why they should try it.

Then the team narrowed the positioning to spreadsheets.

The effect was immediate and measurable. People started paying without being heavily persuaded.

Conversion improved. The message was landing because it was specific enough to be legible.

The decision to build as an AI-native product rather than an AI-enabled one is also worth examining carefully.

Without the intelligence layer, there is no Decide. The AI is not a feature added to a spreadsheet tool. It is the entire product.

That distinction matters for how the company approaches reliability.

Every iteration is about making the AI more accurate, more contextually aware, and more capable across a wider range of spreadsheet tasks.

There is no fallback to a non-AI version of the product. That commitment focuses the roadmap in a way that AI-enabled products with legacy functionality cannot replicate.

The workflow embedding strategy is the third decision worth examining.

Rather than waiting for users to come to Decide’s platform, the product roadmap is explicitly designed to go to where users already are.

The upcoming Decide Mail feature means users can analyse spreadsheet attachments directly inside their email without switching context.

The Excel-native agent and the planned Google Sheets integration mean users can access Decide from within the tools they use every day.

That is not a feature expansion. It is a distribution strategy. The most powerful AI tools are the ones that become invisible because they live inside the workflow rather than beside it.

Ecosystem Context

Decide is built in Lagos by a Nigerian founder for a global user base. It is incorporated in Delaware.

The founder describes this as a strategic decision, enabling access to investment from anywhere in the world and providing a more universally recognised legal structure for institutional and enterprise relationships.

That choice carries a cost that is less often named. Filing requirements, registered agent fees, compliance obligations, and the administrative burden of maintaining a foreign entity from Nigeria without local support infrastructure are real operational distractions.

The founder acknowledges the friction. He accepts it because the alternative, operating only under a Nigerian entity, may limit both the investor conversations available to him and the enterprise credibility.

The data privacy question is an open and growing consideration for any AI tool that ingests professional spreadsheet data.

The founder acknowledges that users uploading sensitive business data to a young platform with a small team will reasonably question how that data is handled.

He describes working toward certifications that will make users more confident. That work is not yet complete.

In a category where enterprise adoption requires security and compliance infrastructure that takes time to build, the gap between the product’s current capability and the trust architecture required for institutional scale adoption is a real consideration.

Observed Patterns

The founder’s previous engineering disciplines, working under production load at scale with zero tolerance for silent failure, are directly relevant to building an AI agent that professionals depend on for work with real consequences.

The speed advantage the founder describes as a competitive moat is also credible in a specific context.

As competitors prioritise the spreadsheet use case and allocate engineering resources toward it, the speed advantage may narrow.

The window in which the small team advantage is most valuable is precisely the current period, before the category is crowded with well-resourced competitors.

The self-awareness around what he would do differently, narrowing the positioning earlier, choosing spreadsheets from the start rather than data analysis, reflects exactly the kind of honest learning that makes the second major decision in any company better than the first.

He has documented this lesson explicitly and articulated it as advice to other founders. Founders who can do that from their first company rather than their second or third tend to avoid the same mistake twice.

Open Variables

The trust architecture for enterprise and professional adoption is the most consequential open variable in the near term.

High value professionals are already using the platform. That is a meaningful organic validation signal.

But informal individual use by professionals at large organisations is different from contracted enterprise adoption.

The latter requires data processing agreements, security certifications, compliance audits, and procurement approval processes that has not yet completed.

The path from individual professional users at enterprise companies to enterprise contracts with those same companies requires building the trust infrastructure that makes the procurement conversation possible.

How quickly that can be built alongside product development is an open question.

The revenue scale relative to infrastructure cost is described honestly by the founder as close enough to not be a major concern, but not yet sufficient to fully self-fund operations.

At pre-seed stage with 21,000 spreadsheets processed and thousands of users, the unit economics of serving AI compute costs per analysis against subscription revenue per user will become more important as scale increases.

Whether the pricing model holds at ten times the current volume is an open variable that the current data does not yet answer.

Why This Matters

For investors evaluating AI productivity tools from emerging markets, the Decide case presents a specific and underexamined opportunity profile.

The founder has deep technical credibility from a high-stakes production engineering background.

The product has achieved benchmark-level accuracy in a specific domain. The user base already includes professionals from top-tier global firms paying voluntarily without being aggressively sold to.

The primary gaps, trust infrastructure, distribution at scale, and the capital to build both, are exactly the gaps that pre-seed investment exists to close.

For ecosystem operators interested in the global knowledge worker productivity market, the observation that one billion Excel users have no specialist AI agent built specifically for their primary tool is the clearest version of a category gap.

The market exists. The professional need is documented and daily. The accuracy threshold required for professional trust has now been demonstrated.

The question is which company builds the distribution to match the capability.

Final Strategic Takeaway

The category of people who work with spreadsheets professionally, daily, in consequential ways, is enormous and global and largely untouched by tools designed specifically for them.

But the path taken by the founding team in developing Decide is clearly drawn as follows:

• Narrow the positioning until it is impossible to misunderstand.
  
• Build the accuracy until professionals cannot afford to not trust it.
  
• Embed the product in the workflows where the users already live.
  
• Become inevitable before becoming famous.
  

That sequencing, product discipline before distribution ambition.

Whether the resources arrive in time to capitalise on the window before well-funded competitors focus on the same category is the open question that the next twelve months will begin to answer.

---

This article is drawn from an in-depth founder interview conducted by Afriq IQ with Abiodun Adetona, founder of Decide. Selected insights and observations are published here.