Nigerian AI Startup Decide Ranks Fourth Globally for Spreadsheet Accuracy

By Adedapo Adesanya

Nigerian startup, Decide, has emerged as the fourth most accurate Artificial Intelligence (AI) agent for spreadsheet tasks globally, according to results from SpreadsheetBench, a widely referenced benchmark for evaluating AI performance on real-world spreadsheet problems.

According to the founder, Mr Abiodun Adetona, the ranking places Decide alongside well-funded global AI startups, including Microsoft, OpenAI, and Anthropic.

Mr Adetona, an ex-Flutterwave developer, also revealed that Decide now has over 3,000 users, including some who are paying customers, a signal to the ability of the startup to scale in the near future.

SpreadsheetBench is a comprehensive evaluation framework designed to push Large Language Models (LLMs) to their limits in understanding and manipulating spreadsheet data. While many benchmarks focus on simple table QA, SpreadsheetBench treats a spreadsheet as a complex ecosystem involving spatial layouts, formulas, and multi-step reasoning. So far, only three agents rank higher than Decide, namely Nobie Agent, Shortcut.ai, and Qingqiu Agent.

Mr Adetona said SpreadsheetBench measures how well AI agents can handle practical spreadsheet tasks such as writing formulas, cleaning messy data, working across multiple sheets, and reasoning through complex Excel workflows. Decide recorded an 82.5% accuracy score, solving 330 out of 400 verified tasks.

“The result reflects sustained investment in applied research, product iteration, and learning from real-world spreadsheet workloads across a wide range of use cases,” Mr Adetona told Business Post.

For Mr Adetona, who built Decide out of frustration with how much time professionals spend manually cleaning data, debugging formulas, and moving between sheets, “This milestone highlights how focused engineering and domain-specific AI development can deliver frontier-level performance outside of large research organisations. By concentrating on practical business data problems and building systems grounded in real user environments, we believe smaller teams can contribute meaningfully to advancing applied AI.”

“For Decide, this is a foundation for continued progress in intelligent spreadsheet and analytics automation,” he added.