# Deductive AI

Aiolum story: https://aiolum.com/story/deductive-ai/ · added 2026-11-08 · updated 2026-11-08

**In short.** Deductive AI (AI SRE: agents that investigate production incidents and find their root cause, United States): Sold to Elastic after about US$1M ARR reported, August 2026. Evidence grade D: reported, not verified. Team: 2 co-founders (Rakesh Kothari and Sameer Agarwal); the About page listed 9 people in November 2026; headcount at the sale not published. Launched 12 November 2025; sold 9 months after launch.

AI SRE: agents that investigate production incidents and find their root cause · United States · AI-native · team of 2 to 10 · sold · group: Developer tools and coding agents

- Outcome: Sold to Elastic after about US$1M ARR reported, August 2026 (evidence grade of the headline number: D; headline in US dollars: 1000000)
- Launch: 12 November 2025; months from launch to the outcome: 9
- Team: 2 co-founders (Rakesh Kothari and Sameer Agarwal); the About page listed 9 people in November 2026; headcount at the sale not published
- Outside funding: US$7.5M seed round led by CRV, announced 12 November 2025; bought by Elastic, completed 24 August 2026
- Main factor: Team
- Meters, 1 to 4: evidence strength 1, copyable pattern 1, capital needed 3
- Tags: AI-native, Small team, Funded, Exit, B2B, Developer tools

Links:

- Visit deductive.ai: https://www.deductive.ai/

**The product.** An AI agent for the engineers who keep production software running. When an alert fires, the agent connects the customer's code, logs, metrics, traces and internal notes, forms hypotheses about what broke, tests them against the data and shows the evidence. Engineers can also ask it questions in Slack, Microsoft Teams or their code editor. The company's pricing page listed a Starter plan at US$999 a month in April 2026, and Enterprise at a custom price.

**Why it worked.** The two founders had spent their careers building data systems at Databricks and ThoughtSpot, and they said that half of their engineering time went to debugging. They built an agent that does the search across logs and code that an on-call engineer does by hand, and they named DoorDash and Foursquare as production users when they came out of stealth in November 2025. AI Chat Daily reported in June 2026 that Elastic agreed to buy the company; Elastic announced the deal in July and completed it in August 2026.

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; E = no revenue figure found, so the headline is a non-money fact and the other facts carry their own grades; calc = computed on this page from graded facts.

## Summary

Deductive AI sold an AI SRE. SRE means site reliability engineering, the job of keeping production software running and fixing it when it fails; an AI SRE is software that does part of that job. When an incident starts (an outage, an error spike or slow responses that set off an alert), Deductive's agents read the customer's telemetry, the logs, metrics and traces that running software records about itself, together with the source code and the team's notes. They look for the root cause, the first change or fault that set off the chain of failures, and present it with the diffs, traces and metrics as evidence. The company's pricing page in April 2026 listed a Starter plan at US$999 a month with 1,000 quick and 10 deep investigations, and Enterprise at a custom price.

Rakesh Kothari, an early engineer at ThoughtSpot, and Sameer Agarwal, who created the BlinkDB research system at UC Berkeley and was one of the first engineers at Databricks, started the company. AI Chat Daily wrote that it was founded in 2023; an Internet Archive copy shows the domain for sale in January 2022 and a Deductive AI product page in April 2024. It came out of stealth (the period in which a startup works without public announcements) on 12 November 2025, when the company announced a seed round, its first institutional funding, of US$7.5 million led by CRV, and published case studies with DoorDash, Foursquare and Kumo AI. Its About page lists the two co-founders and seven other people.

On 19 June 2026 AI Chat Daily reported, citing a person with knowledge of the deal, that Elastic had agreed to buy Deductive for up to US$85 million and that Deductive had reached about US$1 million of ARR (annual recurring revenue: the yearly value of the subscriptions running at one moment). Elastic, a public company, announced the agreement on 22 July 2026 and completed the purchase on 24 August 2026 without stating a price. So the sale is confirmed by the buyer (grade A), while the ARR and the price rest on one anonymous source (grade D). "Up to" means that the full amount is not fixed; the article does not say which part was paid at closing and what the rest depends on.

## Pricing

Deductive AI prices on 16 April 2026, read from an Internet Archive copy of the company's pricing page (grade C). On 8 November 2026 the live pricing page answered "Page Not Found", after the sale to Elastic

| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Starter |  | 999 | 1,000 quick and 10 deep investigations, unlimited users, 1 workspace |
| Enterprise | custom | custom | limits set by incident volume, unlimited workspaces, BYOC, SSO, SLA |

The Starter plan was sold month to month with "Cancel anytime"; the page gave no yearly price. Both plans included integrations with code editors (Cursor and Claude Code through MCP, the Model Context Protocol that lets an AI tool call outside services), Slack and Microsoft Teams. Enterprise added higher investigation limits, unlimited workspaces, single sign-on and role-based access, a service-level agreement, and bring your own compute (BYOC), where the agent runs in the customer's own cloud account.

## How much did Deductive AI make, and what was it sold for? The numbers, graded

A two-founder AI SRE company that raised a seed round, came out of stealth in November 2025 and was bought by Elastic nine months later; the sale is confirmed by the buyer, the revenue and the price by one anonymous source

| Fact | Value | Grade | Note | Sources |
|---|---|---|---|---|
| Annual recurring revenue before the sale | about US$1 million | D | AI Chat Daily, 19 June 2026, by Jaeden Schafer: "DeductiveAI had reached approximately $1 million in annual recurring revenue before the sale, per the same source", the source being "a person with knowledge of the deal". Elastic and Deductive did not respond to the article, and no second source gives a revenue figure. | [1] |
| Reported price | up to US$85 million | D | AI Chat Daily, 19 June 2026: Elastic agreed to acquire DeductiveAI "for up to $85 million, according to a person with knowledge of the deal". The article does not say how much of it is fixed and what the rest depends on. Elastic's release, blog post and quarterly results give no price. | [1, 2, 3, 4] |
| Sale | agreement announced 22 July 2026; completed 24 August 2026 | A | Elastic's blog post: "Elastic announced that it has entered into an agreement to acquire Deductive AI". Elastic's release of 24 August 2026: "Elastic (NYSE: ESTC) has completed the acquisition of Deductive AI". Its results for the first quarter of fiscal 2027 (ended 31 July 2026) list "Acquired Deductive AI" under other business highlights. | [3, 2, 4] |
| Customers at the sale | existing customers, number not stated | A | Elastic: "Existing Deductive AI customers will continue to receive support while integration plans are developed." | [2, 3] |
| Reported price compared with the reported ARR | up to about 85 times | calc | Calculated: US$85 million divided by US$1 million. Both numbers are the grade D figures above; the multiple is an upper bound because the price is "up to". | [1] |
| Reported price compared with the seed valuation | up to about 2.6 times | calc | Calculated: US$85 million divided by the US$33 million seed valuation that AI Chat Daily reported (2.58). The article itself writes "roughly 2.6x its last private mark". | [1] |
| Seed round | US$7.5 million, led by CRV | C | Deductive's launch post, 12 November 2025: "led by Max Gazor at CRV, with participation from Databricks Ventures, Thomvest Ventures, and Primeset", plus angel investors including Ion Stoica, Ajeet Singh and Abhinav Asthana. AI Chat Daily repeats the amount. | [5, 1] |
| Seed valuation | US$33 million | D | AI Chat Daily only. The company's own post gives no valuation. | [1] |
| Founded | 2023 | D | AI Chat Daily: "founded in 2023". The Internet Archive copy of deductive.ai of 13 January 2022 is a Dan.com page, "The domain name deductive.ai is for sale", so no earlier product used the domain. The first readable product page in the archive is from 18 April 2024. | [1, 13] |
| Public launch | 12 November 2025, out of stealth | C | Launch post: "We are excited to finally share what we have been building and to announce our launch out of stealth." It says the product was "already running in production". | [5, 9] |
| Founding team | 2 co-founders: Rakesh Kothari (CEO) and Sameer Agarwal (CTO and CPO) | C | About page: "We (Rakesh and Sameer) worked at two pioneering companies, Databricks and ThoughtSpot, as early engineers". Elastic's blog post welcomes "cofounders Rakesh and Sameer". | [6, 3] |
| People listed on the About page, November 2026 | 9, including the 2 co-founders | C | Four founding engineers, a founding designer, an AI research engineer and a director of go-to-market, read on 8 November 2026. Elastic welcomed "the rest of the Deductive AI team" without a number. | [6, 3] |
| Starter price, April 2026 | US$999 a month, month to month | C | Internet Archive copy of the pricing page, 16 April 2026. Enterprise: custom. | [8] |
| Named production users | DoorDash, Foursquare, Kumo AI | C | Customer stories on Deductive's blog, all dated 12 November 2025, and quotes with names and job titles on the home page. The stories do not say what the customers paid. | [9, 10, 7] |
| Datadog account cut off | 15 December 2025; telemetry restored on Grafana Cloud within 48 hours | C | Deductive's post of 9 January 2026: Datadog put its account "under review" and deactivated its accounts and API keys. Deductive had used Datadog's APM "since February 2025" for its own systems. Datadog sells Bits AI, which the post calls "a direct competitor to Deductive". | [11] |
| Code sandbox for the agent | P99 allocation under 21 ms | C | Deductive's engineering post of 17 August 2026: the agent runs generated code in sandboxes built with gVisor and Cilium, with "P99 sandbox allocation under 21ms and command execution overhead under 10ms". | [12] |
| Company | Deductive AI, Inc., a Delaware corporation | C | Terms of service. Disputes go to courts in San Mateo County, California. | [15] |
| Time from public launch to the completed sale | 9 months | calc | Calculated: 12 November 2025 to 24 August 2026 is 285 days, 9 calendar months. | [5, 2] |

How to read this table: grade A is Elastic's own release, blog post or results statement. Grade C is Deductive's own pages and Internet Archive copies of them. Grade D is a figure reported by one news article that cites one unnamed person. An "X" value was computed on this page from the rows above it.

## When did Deductive AI launch, and how did it grow? Timeline

From a domain for sale in January 2022 to a purchase by Elastic in August 2026. Each entry carries its grade.

- **January 2022** (grade C [13]): The domain deductive.ai is listed for sale on Dan.com, according to an Internet Archive copy.
- **2023** (grade D [1]): Rakesh Kothari and Sameer Agarwal found the company, according to AI Chat Daily.
- **18 April 2024** (grade C [13]): The first readable archived product page promises to "root cause any production incident" and "100x faster MTTR" (mean time to resolve).
- **February 2025** (grade C [11]): Deductive starts to use Datadog's APM for its own production systems.
- **12 November 2025** (grade C [5, 9]): Out of stealth with a US$7.5 million seed round led by CRV, as the company announced, and customer stories with DoorDash, Foursquare and Kumo AI.
- **15 December 2025** (grade C [11]): Datadog reviews and deactivates Deductive's accounts. The team moves its telemetry to an open-source stack on Grafana Cloud in under 48 hours and writes about it on 9 January 2026.
- **16 April 2026** (grade C [8]): The pricing page lists Starter at US$999 a month and Enterprise at a custom price.
- **19 June 2026** (grade D [1]): AI Chat Daily reports, citing a person with knowledge of the deal, that Elastic agreed to buy Deductive for up to US$85 million and that Deductive had about US$1 million of ARR.
- **22 July 2026** (grade A [3]): Elastic announces an agreement to acquire Deductive AI, with no price.
- **17 August 2026** (grade C [12]): Deductive publishes how its agents run generated code in gVisor sandboxes.
- **24 August 2026** (grade A [2]): Elastic completes the acquisition. Existing customers keep their support while integration plans are developed.
- **27 August 2026** (grade A [4]): Elastic's first-quarter results for fiscal 2027 list the acquisition among the quarter's highlights.
- **8 October 2026** (grade A [14]): Elastic announces "Elastic nightshift AI SRE". The headline read for this page does not say whether it uses Deductive's technology.
- **8 November 2026** (grade C [7, 8]): deductive.ai is still online with the banner "Deductive AI is joining forces with Elastic"; its pricing page answers "Page Not Found".

## Why did Deductive AI work?

### Founders from the data systems they debug

Both founders built large data platforms (Apache Spark at Databricks, the search and analytics engine at ThoughtSpot) and had debugged them as engineering leaders. AI Chat Daily wrote that this background explains the price: in its words, the buyer paid for the team and the agent architecture, not the revenue.

### Hypotheses tested against evidence

The agent does not only list signals that changed at the same time. It builds a knowledge graph of services, code and dependencies, proposes causes, checks each one against traces and recent changes, and shows the evidence. In the DoorDash case study it traced a latency spike in an ads API to timeouts from an upstream machine-learning platform that was being deployed at the same time, and the engineers confirmed it in the logs and traces.

### Named users before the announcement

The company came out of stealth with three customer stories and quotes from named engineering leaders at DoorDash and Foursquare. For a product that reads a customer's code and production data, named users in production are the proof that buyers ask for first.

### A buyer that owns the telemetry

An AI SRE needs logs, metrics and traces. Elastic already stores them for its observability customers, so Deductive's investigation engine fits on top of data that Elastic holds. Datadog, the other large owner of such data, had cut Deductive off in December 2025 and sells its own competing agent.

## How it was built

These details come from Deductive's own blog posts and pages (grade C) and Elastic's announcements (grade A).

- A knowledge graph of the customer's codebase, telemetry, engineering discussions and knowledge bases, which several agents query together during an investigation.
- The launch post describes reinforcement learning, training by reward for actions that led to a correct result: the agents learn which investigation steps restored stability in past incidents. Elastic's release quotes its chief executive calling it a "reinforcement learning harness for root cause analysis".
- Generated code runs in sandboxes isolated with gVisor, a user-space kernel, and Cilium network rules that block outgoing traffic by default, with sandbox allocation under 21 milliseconds at the 99th percentile.
- Offered as SaaS or self-hosted; the company says it does not store customer data on its servers or train models for other customers on it.
- Its own telemetry ran on Datadog from February 2025 until Datadog cut it off in December 2025, then on Prometheus, Tempo, Loki and Grafana Alloy with Grafana Cloud as the managed backend. The team used Cursor connected to Deductive through MCP during that move.
- The sources read for this page do not name the language models or the model providers that the agents used.

Sources: [7, 5, 2, 12, 11]

## Who tried this before Deductive AI, and why did they fail?

Three companies in our database of failed startups promised to monitor production software or find errors faster and failed. 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.

### OpTier (2003 to 2015; funding US$112M; cause: Competition)

- Promise: Monitor billions of IT and business transactions a day in real time.
- Why it died: AppDynamics and New Relic moved faster to cloud-native monitoring, and its older architecture could not keep up.
- Record: 

### Proxino (2011 to 2014; funding US$120K; cause: Competition)

- Promise: Real-time error detection and debugging for JavaScript applications.
- Why it died: Sentry and other more complete error tools took the market, and it did not build a developer community.
- Record: 

### Verelo (2012 to 2013; funding US$200K; cause: Ran Out of Cash)

- Promise: Website monitoring for downtime, malware and database connections from many locations.
- Why it died: Larger monitoring services bundled the same checks, and too few businesses paid for a specialised tool.
- Record: 

## What changed

OpTier, Proxino and Verelo collected signals and showed them to people: transactions, errors, uptime checks. The engineer still had to read the dashboards and decide what had caused the problem, and each company competed with larger monitoring vendors that collected the same signals.

Deductive sold the step after the dashboard: the reasoning that an on-call engineer does to find the cause. Language models made it possible for a small team to build agents that read code, logs and traces together, and the founders had built data platforms at Databricks and ThoughtSpot. It also did not have to beat the large vendors: according to AI Chat Daily, one of them paid up to US$85 million for it after it reached about US$1 million of ARR.

## Can one person copy Deductive AI now?

**Verdict: Not for one person without this background.**

Deductive needed two founders with long records in distributed data systems, a seed round of US$7.5 million as the company announced, and enterprise customers willing to connect their production code and data. One person can still build a narrow version: an investigation agent for one stack or one kind of incident, sold to small teams, but the large observability vendors now sell their own agents.

1. Pick one kind of production failure that engineers debug by hand every week, such as failed data pipelines or slow API endpoints.
2. Connect the agent to the code and the telemetry together, and make it show the evidence for each conclusion.
3. Run it in production at a few named companies before any announcement, and publish their stories.
4. Keep the telemetry layer on open standards, so that no single vendor can switch the product off.

**What is still open.** Investigation agents for stacks that the large vendors serve badly: data pipelines, machine-learning training jobs, mobile releases, and companies that cannot send their data to an outside cloud.

## What could go wrong for Deductive AI?

- **What is not verified.** The ARR of about US$1 million and the price of up to US$85 million rest on one article: AI Chat Daily of 19 June 2026, by Jaeden Schafer, which cites "a person with knowledge of the deal"; Elastic and Deductive did not respond to it. The US$33 million seed valuation rests on the same article. Elastic's announcement (22 July 2026), its completion release (24 August 2026) and its quarterly results (27 August 2026) confirm the purchase but give no price, and the article does not say which part of "up to" was fixed. A reader could look for the purchase price in Elastic's quarterly report (Form 10-Q) for the quarter ending 31 October 2026; this page did not read SEC filings.
- **Team size at the sale.** Two co-founders started the company, and it raised outside funding. Its About page lists 9 people, and Elastic welcomed "the rest of the Deductive AI team" without a number. No source read for this page gives the headcount at the sale, so the team is shown as small with the listed count.
- **Competition.** Firetiger, another story on this site, also built agents that investigate production incidents and was bought by Cursor in August 2026. AI Chat Daily names Resolve AI as a funded rival, which it reported at a US$1.5 billion valuation in April 2026; Datadog sells Bits AI, and Elastic itself announced "Elastic nightshift AI SRE" on 8 October 2026.
- **Dependence on platforms and model providers.** Datadog deactivated Deductive's accounts in December 2025, which the company describes as cutting off its own production telemetry within minutes; it moved to an open-source stack on Grafana Cloud. The sources read for this page do not name the language models that the agents used, so the dependence on a model provider cannot be judged from them. The April 2026 pricing page listed integrations with Cursor and Claude Code through MCP.
- **The product after the sale.** On 8 November 2026 the pricing page answered "Page Not Found" while the home page still offered demos. Elastic says existing customers continue to receive support while integration plans are developed; whether new customers can still buy Deductive on its own is not stated.
- **Placeholder text on the site.** The home page testimonials carry leftover template labels ("Advanced User Segmentations", "Banners", "Hints"), and the changelog repeats one entry dated 11 October 2024 four times beside a "1.4.0" entry dated 30 September 2023. This page does not use the changelog dates as evidence.

## Sources

1. AI Chat Daily, 19 June 2026, by Jaeden Schafer: Elastic agrees to buy DeductiveAI for up to $85M in AI SRE land grab — https://www.aichatdaily.com/ai-business/elastic-agrees-buy-deductiveai-up-85m-ai-sre (grade D). Read in full (downloaded by the pipeline's harvester on 8 November 2026). Source for the reported price, the ARR, the seed valuation and the founding year, all from one unnamed person with knowledge of the deal.
2. Elastic press release, 24 August 2026: Elastic Completes Acquisition of Deductive AI — https://ir.elastic.co/News--Events/news/news-details/2026/Elastic-Completes-Acquisition-of-Deductive-AI/default.aspx (grade A). Read in full. The buyer's release on its investor-relations site: completion, the technology, the former CEO's quote and the support for existing customers. No price.
3. Elastic Blog, 22 July 2026, by Baha Azarmi: Elastic and Deductive AI join forces to accelerate agentic incident investigation for engineering teams — https://www.elastic.co/blog/elastic-deductive-ai (grade A). Read in full. The buyer's announcement of the agreement and of the co-founders joining Elastic. No price.
4. Elastic press release, 27 August 2026: Elastic Reports First Quarter Fiscal 2027 Financial Results — https://ir.elastic.co/News--Events/news/news-details/2026/Elastic-Reports-First-Quarter-Fiscal-2027-Financial-Results/default.aspx (grade A). Read the highlights and searched the whole release for Deductive. It lists "Acquired Deductive AI" and gives no price.
5. Deductive AI blog, 12 November 2025: Introducing Deductive: The AI SRE for Fast Moving Teams — https://www.deductive.ai/blogs/introducing-deductive-the-ai-sre-for-fast-moving-teams (grade C). Read in full. The founders' launch post: background, reinforcement learning, the seed round and its investors.
6. Deductive AI, About us — https://www.deductive.ai/about (grade C). Read in full on 8 November 2026. Founding story and the list of 9 team members.
7. Deductive AI home page — https://www.deductive.ai/ (grade C). Read in full on 8 November 2026. Product description, deployment options, customer quotes and the Elastic banner.
8. Deductive AI pricing page, Internet Archive copy of 16 April 2026 — https://web.archive.org/web/20260416103411/https://www.deductive.ai/pricing (grade C). Read in full. The live page answered "Page Not Found" on 8 November 2026.
9. Deductive AI blog index — https://www.deductive.ai/blogs (grade C). Read in full on 8 November 2026. Titles and dates of all posts, including the three customer stories of 12 November 2025.
10. Deductive AI blog, 12 November 2025: How DoorDash Powers a Reliable, High-Performance Ad Platform with Deductive AI — https://www.deductive.ai/blogs/how-doordash-powers-a-reliable-high-performance-ad-platform-with-deductive-ai (grade C). Read in full. Case study of one production incident; no money figures.
11. Deductive AI blog, 9 January 2026: Datadog, Thank You for Blocking Us — https://www.deductive.ai/blogs/datadog-thank-you-for-blocking-us (grade C). Read in full. Source for the Datadog cut-off, the move to Grafana Cloud and the use of Cursor through MCP.
12. Deductive AI blog, 17 August 2026: Building Secure Sandboxes for AI Agents — https://www.deductive.ai/blogs/building-secure-sandboxes-for-ai-agent-execution (grade C). Read the introduction and the sections on the gVisor and Cilium choice.
13. Internet Archive copies of deductive.ai, 13 January 2022 and 18 April 2024 — https://web.archive.org/web/20220113083500/https://deductive.ai/ (grade C). Read both copies and the archive's capture list. The 2022 copy is a domain-sale page; the April 2024 copy (https://web.archive.org/web/20240418095538/https://www.deductive.ai/) is the first readable product page.
14. Elastic investor relations, list of 2026 press releases — https://ir.elastic.co/News--Events/news/default.aspx (grade A). Read the 2026 list through the page's own data feed. Source for the dates and headlines of the Deductive releases and of the nightshift AI SRE announcement; the nightshift release itself was not read.
15. Deductive AI, Terms of Service — https://www.deductive.ai/terms (grade C). Read the sections naming the company and the governing law.
16. Database of failed startups, local export — /method/#dead (grade C). Local export. AI-written summaries, not independently verified. Source for the failed predecessors.

Contact: aiolum@livyden.com. Run Deductive AI? Send corrections or a verified figure to aiolum@livyden.com (subject: Correction: Deductive AI).