# Artificial Workflow (Echobase)

Aiolum story: https://aiolum.com/story/artificial-workflow/ · added 2026-10-31 · updated 2026-10-31

**In short.** Artificial Workflow (Echobase) (AI agents that answer questions from a company's own files, Australia): US$10K a month, June 2024. Evidence grade D: reported, not verified. Team: 3 founders and 2 employees in June 2024, per Starter Story. Launched October 2023; closed later.

AI agents that answer questions from a company's own files · Australia · AI-native · team of 2 to 10 · closed · group: Documents, data and slides

- Outcome: US$10K a month, June 2024 (evidence grade of the headline number: D; headline in US dollars: 120000)
- Launch: October 2023; months from launch to the outcome: 8
- Team: 3 founders and 2 employees in June 2024, per Starter Story
- Outside funding: No outside funding named in any source
- Main factor: Distribution
- Meters, 1 to 4: evidence strength 1, copyable pattern 2, capital needed 2
- Tags: AI-native, Small team, Bootstrapped, Closed, B2B, Information Technology

Links:

- Archived echobase.ai, June 2024: https://web.archive.org/web/20240613043636/https://echobase.ai/

**The product.** Echobase was a web app in which a team uploaded its files, or synced a cloud storage service, and then built AI agents that answered questions about those files, drafted new text from them and found themes across them. It ran on OpenAI's GPT-3.5 and GPT-4 models. It was sold to consultants and consulting firms, and the co-founder said in June 2024 that most of the money came from an Enterprise plan that the pricing page listed at US$299 a month for 5 seats.

**Why it worked.** The founders first sold custom AI projects to consulting firms and noticed that they were building the same document-analysis tool for each client. They packed that tool into one subscription product. One co-founder, a former management consultant at EY, then sold it to consulting firms through his own network, which the founders said brought most of the paying customers.

Evidence grades: A = disclosed by a public company, a filing or a payment processor, or verified by one; B = two or more independent reports agree, or a journalist confirmed it; C = a founder claim that no one else verified; D = reported, not verified: one founder claim or a third-party estimate with no second source; calc = computed on this page from graded facts.

## Summary

Echobase was a retrieval-augmented generation (RAG) product. Retrieval-augmented generation means that, before a language model answers a question, the program searches the user's own documents for the passages that matter and puts them into the model's prompt, so the answer is based on those documents and not only on what the model learned in training. In Echobase a user uploaded files such as interview transcripts or proposals, created an AI agent (here, a chat assistant tied to one set of documents, with its own settings and access rules) and asked it to answer, draft or analyse. The company behind it, Artificial Workflow Pty Ltd, was in Brisbane, Australia.

The co-founders named in the only interview are Andrew, who said his background was marketing in the Web3 industry, and James, a former management consultant at EY; Starter Story lists 3 founders and 2 employees. They began as an AI consulting company, made about 200 sales calls to consulting firms found on LinkedIn, and turned the work they kept repeating into a subscription product. The Echobase site was online by October 2023. In a Starter Story interview published on 17 June 2024, Andrew said that Echobase made roughly US$10,000 a month in recurring revenue, that 30 enterprise customers subscribed, and that most of the money came from the US$299 Enterprise plan; he also said the founders had paid about US$50,000 to developers whose work they later had to replace.

That single figure is the headline of this page and rests on one source, so the page grades it D, reported and not verified. Starter Story's page now notes that the business started in 2023 and ended in 2024, with the reason "Shut down". The Echobase site stayed online with a 2025 copyright under another company name until at least April 2026, and in October 2026 its address showed only a hosting error page. The page also shows what happened to earlier companies that tried to answer questions from a company's own knowledge.

## Pricing

Echobase prices from the archived pricing page of 30 May 2024 (grade C); the product is no longer sold

| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Free |  | 0 | 20 GPT-3.5 queries, 1 agent, 20 pages |
| Individual |  | 19 | 2,100 queries, 5 agents, 10,000 pages |
| Professional |  | 49 | 6,500 queries, 10 agents, 40,000 pages, GPT-4 |
| Enterprise |  | 299 | 25,000 queries, 20 agents, 300,000 pages, 5 seats |
| Custom |  | on request | assisted setup and training |

All prices were in US dollars per month; the page showed no yearly prices. A query was one question to an agent. GPT-4 queries were counted separately because, in the page's words, GPT-4 cost more: 325 a month on Professional and 1,250 on Enterprise. A Custom plan with assisted setup, custom agents and training had no listed price. The co-founder said in June 2024 that onboarding and training were also sold for a flat US$2,500.

## How much revenue did Artificial Workflow (Echobase) make before it closed? The numbers, graded

A document-question tool for consultants that, by one co-founder's account in June 2024, made about US$10,000 a month, and that Starter Story lists as shut down in 2024

| Fact | Value | Grade | Note | Sources |
|---|---|---|---|---|
| Recurring revenue per month, June 2024 | roughly US$10,000 | D | Andrew, co-founder, in the Starter Story interview published 17 June 2024: "Currently, EchoBase AI is making roughly $10,000 a month in recurring revenue, about 80% of which comes from enterprise clients paying $299 / month." One source, one date. | [1] |
| Run rate from the June 2024 figure | about US$120,000 a year | calc | Calculated: US$10,000 times 12. This is the outcomeUsd value of this page. | [1] |
| Enterprise customers, June 2024 | 30 | D | Same interview: "We currently have 30 subscribed enterprises with some paying for extra seats." The same text gives the Enterprise share of revenue as about 80% in one place and 90% in another. | [1] |
| 30 Enterprise plans at the list price | US$8,970 a month | calc | Calculated: 30 times US$299. That is about 90% of the stated US$10,000, so the customer count and the revenue figure agree with each other. | [1, 3] |
| Training and onboarding fee | US$2,500 flat | D | Same interview: 4 hours of calls with training materials. | [1] |
| Money lost on developers | about US$50,000 | D | Same interview: a cheap Fiverr developer built a first version without asynchronous processing, so one client's upload blocked the others; two more developers and then a development agency rebuilt it. | [1] |
| Team, June 2024 | 3 founders and 2 employees | D | The fact box of the Starter Story page. The interview was also looking for a full-stack developer as co-founder and CTO, for equity. | [1] |
| Sales calls before the product | around 200 | D | Same interview, from cold outreach on LinkedIn to consulting firms when the company still sold custom AI projects. | [1] |
| Search traffic, about May 2024 | almost 10,000 clicks in one month | D | Same interview, from blog posts and free tool pages written for search keywords. | [1] |
| Models | OpenAI GPT-3.5 and GPT-4 | C | Archived pricing page, 30 May 2024. Open-source models were listed as "coming soon". | [3] |
| Earliest archived site | 24 October 2023 | C | Wayback Machine copy of echobase.ai with the footer "© 2023 Artificial Workflow Pty Ltd." Starter Story's fact box says the business started in December 2023. | [5, 1] |
| Closure | ended in 2024 | D | Starter Story's note on the interview page: "This business is no longer running. It was started in 2023 and ended in 2024. Reason for closure: Shut down." | [1] |
| Time from launch to the June 2024 figure | 8 months | calc | Calculated: October 2023, the first archived copy of the site, to June 2024. | [1, 5] |

How to read this table: a grade D value is a single founder statement with one date and no second source. A grade C value is read from the company's archived or current web pages. A "calc" value was computed on this page from the rows above it. No value is grade A or B.

## When did Artificial Workflow (Echobase) launch, and how did it grow? Timeline

From custom AI projects for consulting firms to a subscription product and back out of the market, in about one year. Each entry carries its grade.

- **2023** (grade D [1]): Andrew and James start Artificial Workflow as an AI consulting company and make about 200 sales calls to consulting firms found on LinkedIn. They see that they are building the same document tool for each client.
- **24 October 2023** (grade C [5]): The Echobase site is online: "Echobase is how teams query, create, and analyse from their files." Footer: Artificial Workflow Pty Ltd.
- **Late 2023 to early 2024** (grade D [1]): A first version built by a low-cost Fiverr developer signs 4 to 5 enterprise clients, but lacks asynchronous processing. The founders spend three months rebuilding it with a development agency.
- **30 May 2024** (grade C [3, 4]): The pricing page lists Free, Individual at US$19, Professional at US$49 and Enterprise at US$299 a month, plus free tool pages such as an AI paragraph generator for search traffic.
- **17 June 2024** (grade D [1]): Starter Story publishes the interview: roughly US$10,000 a month in recurring revenue, 30 enterprise subscribers, 3 founders and 2 employees, by Andrew's account.
- **2024** (grade D [1]): Starter Story later marks the business as ended in 2024, reason "Shut down". No date or detail is given.
- **15 June 2025** (grade C [6]): The archived Echobase site is still online, now with "© 2025 Core Flow Intelligence" in the footer. No source explains the new name.
- **31 October 2026** (grade C [7, 8]): echobase.ai shows a Framer "Site Not Found" page. artificialworkflow.com presents an AI automation agency with the same 2025 footer and does not mention Echobase.

## Why did Artificial Workflow (Echobase) work?

### The product came out of repeated consulting work

The founders sold custom AI projects first. When the same document-analysis tool came up for client after client, they made it one product with plans. The calls also told them what consultants wanted to upload: interview transcripts that a team needed to read and summarise.

### Sales by one person with a network

Andrew said in the interview that James, who had worked as a management consultant at EY, booked tens of sales demos a week through cold outreach on LinkedIn, and that about 90% of revenue came from enterprise clients that James had onboarded himself. The channel depended on one person's reputation in one industry.

### A high plan for teams

The pricing page put 5 seats, 25,000 queries and the newest models in an Enterprise plan that it listed at US$299 a month. Andrew said 30 firms were on it, which by itself comes to about 90% of the revenue he stated. A few firms at a team price carried the business, and each also could buy training.

### Free tool pages for search

The site published free single-purpose generators, such as an AI paragraph generator and a passive voice checker, each on its own page aimed at one search term, and linked them together. Andrew said this brought almost 10,000 clicks from search in one month, though he also said cold outreach, not search, brought most paying users.

## How it was built

These details come from the co-founder's Starter Story interview (grade D) and the archived Echobase pages (grade C).

- The first version was designed in Figma from feedback in the LinkedIn sales calls, then built by a low-cost developer hired on Fiverr.
- That version had no asynchronous processing, so when one client uploaded files the others had to wait. Two developers and then a development agency rebuilt it over about three months.
- Answers came from OpenAI's GPT-3.5 and GPT-4 models; GPT-4 questions had their own, smaller monthly allowance on the pricing page.
- The marketing site was built and hosted on Framer, a no-code website builder. Code and project management were on GitHub.
- Mailchimp sent the email campaigns, Ahrefs was used to find search keywords, Xero ran the accounting, and an affiliate program through Rewardful offered partners a 40% cut, per the interview and the site.
- Andrew said the founders would start again by finding a technical co-founder who could build the whole first version, instead of paying outside developers.

Sources: [1, 2, 3]

## Who tried this before Artificial Workflow (Echobase), and why did they fail?

Three companies in our database of failed startups tried to answer questions from a company's own knowledge or to deflect questions with software before the current language models. The values below are from that database and are not independently verified.

The failed startups come from our database of dead startups, retold here in our own words; Its entries are AI-assisted summaries of public sources, so they carry no evidence grade.

### Niles (2017 to 2020; funding US$2.8M; cause: Competition)

- Promise: A chatbot in Slack that held a company's knowledge and answered staff questions.
- Why it died: Slack added its own AI features, and cheaper and faster competitors arrived.
- Record: 

### Answerly (2013 to 2016; funding US$550K; cause: Product/Tech Failure)

- Promise: FAQ chatbots that answered routine support questions so people handled only the hard ones.
- Why it died: Natural language processing was too weak at the time, and Zendesk and Intercom already owned the customers.
- Record: 

### Kippt (2012 to 2017; funding US$400K; cause: Competition)

- Promise: A tool to save, organise and share web content against information overload.
- Why it died: Pocket and Evernote held the market, and the social features did not spread.
- Record: 

## What changed

Niles and Answerly had to build their own language understanding, and the database says it was not good enough. By 2023 Echobase could send a user's documents and question to GPT-4 and get a usable answer, so a small team without machine-learning staff could sell the same promise. That is the change that made the business possible.

The same change removed the barrier for everyone else. Kippt and Niles died when larger products added the feature, and the Echobase comparison pages named ChatGPT, Copilot, Chatbase and several more document-chat products as rivals. Starter Story reported the business as ended within months of the interview in which the co-founder stated US$10,000 a month; no source says whether competition was the reason.

## Can one person copy Artificial Workflow (Echobase) now?

**Verdict: The path into the product, yes. The product itself, with care.**

Selling custom projects first, and turning the part that repeats into a subscription, is open to any small team with access to an industry. A general tool for asking questions about files is not: the model makers and office suites now include it, and the sources say this business stopped within a year of the co-founder's revenue claim.

1. Sell custom AI work to one industry where a founder has worked and has contacts.
2. Write down which parts every client asks for, and build only those into a product with a team plan.
3. Pay for solid engineering from the start, or bring in a technical co-founder; the founders said cheap development cost them about US$50,000.
4. Keep the work that only an industry insider can do, such as setup and training, as part of the offer.

**What is still open.** Narrow tools for one profession's documents, where the setup, the data rules and the training matter more than the model, and where a general chat product does not fit the daily work.

## What could go wrong for Artificial Workflow (Echobase)?

- **What is not verified.** The headline figure, roughly US$10,000 a month in recurring revenue, rests on one source: the Starter Story interview published on 17 June 2024, in which the co-founder Andrew stated it. The 30 enterprise customers, the team of 3 founders and 2 employees and the US$50,000 lost on developers come from the same interview. The interview gives the enterprise share of revenue as about 80% in one sentence and 90% in another. No payment processor, journalist or buyer has confirmed any of it. A reader could look for a revenue profile on a payment-verified site or a second interview with the founders; this page found neither.
- **Why it closed.** Starter Story's note on the interview page says: "This business is no longer running. It was started in 2023 and ended in 2024. Reason for closure: Shut down." It gives no month and no reason beyond that. The record is not clean: archived copies show the Echobase site still online on 15 June 2025 and 8 April 2026, with the footer "© 2025 Core Flow Intelligence" instead of Artificial Workflow Pty Ltd, and no source explains that name or says whether the product still took payments. On 31 October 2026 echobase.ai returned a Framer "Site Not Found" page. So this page treats the business as closed in 2024, on Starter Story's word, and the site as gone by October 2026.
- **The domain now sells agency work.** On 31 October 2026 artificialworkflow.com described an AI automation agency in Brisbane that builds workflows for clients, and does not name Echobase or the founders. No source says whether it is the same team. The page here is about the Echobase software product, not the agency.
- **The founders' own sales method.** Andrew said that most paying customers came from James's cold outreach on LinkedIn, which booked tens of demos a week. The interview describes personal outreach by a known consultant; it does not mention automation tools or bulk messaging. The product itself did not send messages on LinkedIn.
- **One industry, one seller.** About 90% of the revenue, by the interview, came from enterprise clients that one co-founder had onboarded himself. A business whose sales rest on one person's network stops growing when that person stops selling.

## Sources

1. Starter Story, 17 June 2024: We Created AI Tools For Consultants And Make $10K/Mo — https://www.starterstory.com/stories/artificial-workflow (grade D). Read in full. A founder interview with Andrew, co-founder, not a page marked "Estimated from public sources". Source for the revenue figure, the customer count, the team, the build history, the sales method, the tools and the closure note.
2. Wayback Machine copy of echobase.ai, 13 June 2024 — http://web.archive.org/web/20240613043636/https://echobase.ai/ (grade C). Read in full. Source for the product description, the free tool pages, the affiliate program and the footer naming Artificial Workflow Pty Ltd.
3. Wayback Machine copy of the Echobase pricing page, 30 May 2024 — http://web.archive.org/web/20240530042721/https://echobase.ai/pricing (grade C). Read in full. Source for the plans, prices, query limits and the GPT-3.5 and GPT-4 models.
4. Wayback Machine copy of the Echobase AI Tools page, 30 May 2024 — http://web.archive.org/web/20240530033243/https://echobase.ai/tools (grade C). Read in full. Source for the free generator pages written for search traffic.
5. Wayback Machine copy of echobase.ai, 24 October 2023 — http://web.archive.org/web/20231024221720/https://echobase.ai/ (grade C). Read in full. The earliest full copy in the Wayback Machine index. Source for the launch month used on this page.
6. Wayback Machine copy of echobase.ai, 15 June 2025 — http://web.archive.org/web/20250615121134/https://echobase.ai/ (grade C). Read in full. Source for the site still being online in 2025 with the footer "© 2025 Core Flow Intelligence". A copy of 8 April 2026 shows the same footer.
7. echobase.ai, read 31 October 2026 — https://echobase.ai/ (grade C). Read in full. The address returned HTTP 404 with a Framer "Site Not Found" page.
8. Artificial Workflow home and About pages, read 31 October 2026 — https://artificialworkflow.com/about (grade C). Read in full. The domain now presents an AI automation agency in Brisbane with the footer "© 2025 Core Flow Intelligence"; Echobase is not mentioned.
9. Database of failed startups, local export — /method/#dead (grade C). Local export. AI-written summaries, not independently verified. Source for the failed predecessors.