Cognosys
Last reviewed
Jun 4, 2026
Sources
15 citations
Review status
Source-backed
Revision
v1 · 1,827 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Last reviewed
Jun 4, 2026
Sources
15 citations
Review status
Source-backed
Revision
v1 · 1,827 words
Add missing citations, update stale details, or suggest a clearer explanation.
Cognosys was a Vancouver-based artificial intelligence startup, founded in 2023 by Sully Omar (Sulaiman Omar Marouf) and Homam Malkawi, that built browser-based autonomous AI agents for non-technical users. It launched in April 2023 during the first wave of AutoGPT-style projects, offering a no-code web interface in which a user typed a high-level goal and the system planned and carried out the work by spawning a chain of task-completing agents. In October 2024 the company pivoted and rebranded to Ottogrid, a market-research tool built around a spreadsheet-style "smart table" interface. Ottogrid was acquired by the Canadian foundation-model company Cohere on 16 May 2025, after which the standalone product was wound down and the team's work folded into Cohere's enterprise application North. The original cognosys.ai site no longer runs as a live product.
Cognosys is often cited as an early consumer-facing example of agentic AI: software that decomposes an objective into sub-tasks and executes them with limited human supervision, rather than answering one prompt at a time the way a chatbot does. Its arc, from a free-wheeling "do anything" autonomous agent to a focused enterprise data tool absorbed by a larger lab, mirrors the broader 2023 to 2025 shift in how the industry productized large language model-powered agents.
Cognosys was created in early 2023 by Sully Omar and Homam Malkawi and based in Vancouver, British Columbia. Omar served as chief executive officer; before the company he had worked as a software engineer (including at the website builder Universe) and studied at Simon Fraser University. The product, hosted at cognosys.ai, first appeared publicly around 14 April 2023, in the same weeks that Significant Gravitas's AutoGPT and Yohei Nakajima's BabyAGI popularized the idea of looping an LLM to pursue a goal autonomously.
The pitch was to make that capability usable without code. Where AutoGPT and BabyAGI were developer tools run from a terminal, Cognosys wrapped the same loop in a clean web app: a user entered an objective such as "compile a comprehensive report on Tesla's market performance and future outlook," and the agent broke the goal into a sequence of tasks, worked through them (often by searching the web), and returned a synthesized result. The company described the system as creating "a series of agents each responsible for completing particular tasks," and as showing the intermediate computing steps rather than only a final answer. Output could be returned as formatted text, tables, or code. In its earliest form the web app required the user to supply an OpenAI API key, so it was not entirely free to run.
Over the following months Cognosys grew into a broader productivity layer. It added support for multiple underlying models (over its lifetime these included OpenAI's GPT-3.5 and GPT-4 family and Google's Gemini), connections to outside apps such as Notion, and the ability to schedule recurring agents, for example a market-trend summary or a digest of the morning's news delivered automatically each day. The marketing tagline was "Simplify your workflows with AI." A tiered subscription model accompanied this version: a free plan, a Pro plan around US$15 per month with higher message, integration, and workflow limits, and an Ultimate plan around US$59 per month for effectively unlimited use.
In May 2023 Cognosys announced a seed round of CAD $2 million to build out its AI agents. The round was led by Sangeen Zeb, a general partner at GV (Google Ventures). It drew an unusually founder-heavy roster of angel investors for a company of its size, including Guillermo Rauch, chief executive of Vercel; Amjad Masad, chief executive of Replit; Cohere co-founders Aidan Gomez (chief executive) and Ivan Zhang (chief technology officer); Yohei Nakajima, the creator of BabyAGI; and the fund Untapped Capital (Untapped VC). The company said the capital would go toward expanding the team and refining the platform, and that it planned to broaden model support beyond OpenAI to include Google's PaLM API, Cohere's API, and Anthropic's Claude models. The funding round is the company's only publicly reported financing; there is no evidence that Cognosys participated in Y Combinator or another startup accelerator.
In October 2024 the company overhauled its product and rebranded from Cognosys to Ottogrid. The shift was strategic, not cosmetic. Rather than continue as a general-purpose "do anything" agent, the team narrowed in on automating research and structuring messy data, a use case its founders had seen resonate most strongly with users.
The defining change was the interface. Ottogrid replaced the conversational agent with a "native table interface," a spreadsheet in which each column functions as its own specialized AI agent driven by a prompt. A user supplies a list of items as rows (companies, people, properties, documents) and configures column prompts to extract or compute a value for each one, so many agents run in parallel to fill the grid. The company positioned this for tasks like company research (identifying a chief executive, recent funding, or news), sales-lead enrichment (contact details, decision-makers, technology stacks), real-estate analysis, recruiting, and contract or document review. The transition was announced on Ottogrid's blog under the title "Cognosys is Evolving: Welcome to Ottogrid." Existing Cognosys users were told the old product was being deprecated: they were given a window to download any data they had stored, after which login access would be removed and active subscriptions cancelled, and they were offered a discount to move to Ottogrid.
The company's open-source footprint from this period reflects the data-extraction focus. Its GitHub organization (which uses both the CognosysAI and Ottogrid names) hosts projects in TypeScript and Rust for headless-browser automation, converting web pages into LLM-friendly markdown, redacting personally identifiable information, and agentic retrieval-augmented generation.
On 16 May 2025, Ottogrid co-founder Sully Omar announced on X that the company had been acquired by Cohere, a Toronto-based developer of enterprise foundation models. Financial terms were not disclosed. Reporting on the deal described Ottogrid as a Vancouver platform for automating high-level market research, built on AI-powered document analysis and the native-table interface.
Cohere said it would integrate Ottogrid's "smart table" technology directly into North, its enterprise application for knowledge workers that summarizes documents, extracts insights, and automates research-heavy workflows. Cohere chief executive Aidan Gomez, who had been an angel investor in the company two years earlier, said he was "super excited to bring the Ottogrid team aboard" and to give enterprises "a new way to tackle research with smart tables." Co-founder Homam Malkawi wrote that the team would be "joining the team at Cohere and integrating Ottogrid into their North platform," calling it "the right home for what we've built." As part of the transition the standalone Ottogrid product was to be sunset, with the founders promising customers ample notice, a reasonable transition period, and continued support during the wind-down.
In its 2023 to 2024 form, Cognosys followed the task-driven autonomous-agent loop popularized by AutoGPT and BabyAGI:
This loop-and-decompose design is what distinguished Cognosys from a standard chatbot like ChatGPT, which generates a response to each prompt but does not, on its own, plan and carry out a multi-step task. CEO Sully Omar, in a later interview, said the agents typically ran for anywhere from 60 seconds to about five minutes per task, far longer than the sub-second latency users expect from conventional software, and that the most popular use was a research agent, comparable in spirit to Perplexity, that searched the internet and returned an answer with source links. He also described moving the backend off short-lived serverless functions onto longer-running Google Cloud Run instances to accommodate these multi-minute agent runs.
Cognosys drew attention in spring 2023 as one of the more approachable entries in the AutoGPT moment, frequently shared as a way to "run your own AutoGPT in the browser" without installing anything or writing Python. It was widely catalogued in AI-tool directories and is often grouped with other early autonomous-agent web apps of that period such as AgentGPT.
Its longer significance is as a case study in how 2023's open-ended autonomous agents were productized. The company found that a general "do anything" agent was hard to make reliably useful, and narrowed to a concrete, repeatable job, enriching lists and datasets at scale, before being absorbed by a larger lab building enterprise agent infrastructure. The founders framed the journey as emblematic of the field's move from generative AI toward agentic AI. As of 2026 the cognosys.ai domain no longer serves a working product; visitors land on a paused deployment page, and the brand effectively ended with the rebrand to Ottogrid and that product's subsequent sunset inside Cohere. (A number of "review" pages published in 2026 still describe Cognosys as a live, subscription-priced workflow agent; these appear to recycle pre-rebrand marketing material and do not reflect the product's actual status.)