The short answer: for an institutional desk that wants a pre-read before every print in its coverage and a review after it, AllMind AI is the system built for that run. Its automations hang off the earnings calendar with a lead time you set, read filings, transcripts, entitled broker research and the firm's own model in one pass, and deliver a Word or PDF brief. Quartr Automations (August 24, 2026) is the pick for teams that live in company-published material across 60-plus markets. Marvin Labs is the self-serve option with filing and transcript triggers, and Claude Managed Agents suit a team that will build its own.

Who this is for: analysts, portfolio managers and heads of research at hedge funds, asset managers and sell-side desks, plus investor-relations teams, deciding in 2026 which earnings work can run without a person pressing the button.

Published August 28, 2026. Last reviewed August 28, 2026. Written by the AllMind AI research team.

Reviewed by Anwaar Malik, founder of AllMind AI.

Disclosure: AllMind AI builds one of the eight tools compared here. Where a rival tool handles a trigger better, the section says so; no vendor paid for placement.

Key takeaways

• Four vendors now document an earnings trigger. AllMind AI (calendar trigger with chosen lead time), Quartr (August 24, 2026), Marvin Labs (July 27, 2026) and Anthropic's Claude Managed Agents (nightly schedule, May 5, 2026) each describe research that runs without a person starting it.
• The trigger is the cheap part; the sources are the expensive part. Quartr and Marvin Labs fire on first-party documents. Of the eight, only AllMind AI and AlphaSense read entitled broker research inside the run, and only AllMind AI joins a warehouse table to it.
• Company caps are rarely published. AllMind AI states up to 200 companies per automation on its comparison pages; AlphaSense caps concurrent agents at 10; Quartr, Marvin Labs and Aiera publish no cap.
• Walmart's Q2 FY2027 print shows what the output has to contain. Revenue of $187.9 billion, e-commerce up 23 percent, and a fiscal-year outlook raised to 4.0 to 5.0 percent net sales growth on August 20, 2026, against "unchanged" on May 21, 2026.
• Unattended output is a draft. AllMind AI's verification pass leaves a field blank when it cannot source the figure, Marvin Labs bills credits per run, and every tool here hands the rating and the narrative back to the analyst.

What is an AI agent that runs automatically when a company reports earnings?

An AI agent that runs automatically when a company reports earnings is a saved research task bound to an event on that company's calendar: the earnings date, the moment the release or transcript is published, or the 10-Q filing. When the event fires, the agent reads a defined set of sources, writes a defined output, and sends it to a defined person without anyone opening the platform.

Two dates matter per company: the expected report date, which sets the pre-read, and the actual publication, which sets the review. A tool with only the second date writes nothing before the print; a tool with only the first cannot react when a company moves its date.

Clock triggers vs event triggers: which research tasks belong to each?

Clock triggers suit the work that repeats on your schedule; event triggers suit the work that repeats on the company's. The table sorts the common earnings tasks between them.

AllMind AI, Quartr and Marvin Labs offer both trigger types in 2026. AlphaSense (Custom Workflow Agents) and Claude Managed Agents offer the clock only. Hebbia Max runs on ask. Daloopa and Aiera are source layers whose trigger is the arrival of the filing or the call itself.

What are the best tools for scheduled research automations in 2026?

AllMind AI is the best fit for a desk that wants one automation to cover its coverage list, read entitled and internal sources, and emit a document. The seven others each do one part of that job well. The table sorts them by trigger, sources, output, cap, internal-data route and pricing basis, with a limitation for each.

How we evaluated

Each tool was scored on five checkable questions from its own documentation, opened between August 24 and August 28, 2026: what fires the run, what it may read, what it emits, how many companies one automation covers, and whether internal firm data enters the run. Funding and estimated prices did not move a score; where a vendor publishes no answer, the cell says so.

AllMind AI

AllMind AI runs earnings work as automations built in Agent Studio: recurring research on a schedule, and monitoring agents that watch incoming filings, transcripts and news against a thesis. Its hedge-fund page states the earnings case: a covered name reports, a filing lands, or a company turns up on a conference agenda, and the output is in the inbox. The trigger is keyed to the earnings calendar with a lead time you choose, so the prep for Thursday's print is written on Monday. The sources under a run are S&P, FactSet, LSEG and MSCI data, broker research under the firm's entitlement, Expert Insights transcripts, filings, live calls and news.

Where it wins: coverage-wide runs that need more than public documents. One automation tracks event triggers across up to 200 companies, delivered as Word or PDF, per AllMind AI's comparison pages, and the market-cap floor, the sector map and the send hour live on the automation itself. The firm's Snowflake, Databricks or S3 tables are read in place, without a copy, so the house estimate for Walmart's e-commerce growth sits beside the printed 23 percent. The ontology is the mechanism: the release attaches to the company object and the margin thesis on it, so a broker revising an estimate after the print flags every comp table built on the old number.

Where it falls short: unattended runs are where the verification pass shows its edge case. When it cannot source a figure, the field ships blank, so a release-day brief on a company with a late transcript will carry gaps that a person fills. Live embargoed broker notes need the firm's own RMS entitlement; without it, notes arrive on the aftermarket delay, which varies by broker, so the T+1 roll-up can be thin. There is no self-serve plan, so a retail user or an individual outside an institution is better served by Marvin Labs or Quartr's app.

Quartr Automations

Quartr Automations launched on August 24, 2026 for Quartr Pro users, who can set up research tasks that run on a schedule or trigger the moment earnings and new documents are released. The press release words the scope as all first-party information from public companies, and Quartr's own pages put that at roughly 15,000 to 16,000 companies across 60-plus markets (the home page and the Pro page print different counts on the same day). Results arrive in chat, with optional notifications by email, the activity feed or mobile push, and a Quartr MCP is included with Pro.

Where it wins: the fastest documented event trigger on the widest first-party universe in this table. A global desk covering a Swedish industrial, an Indian bank and a US retailer gets the same trigger on all three.

Where it falls short: the run reads only what the company published. No broker notes, no expert-call transcripts, no house model, and no view of the firm's own documents. Pro and API pricing is contact-sales only as of August 2026, and no per-automation company cap is published.

Marvin Labs scheduled Deep Research Agents

Marvin Labs turned its Deep Research Agents (launched September 3, 2025) into scheduled agents on July 27, 2026. A schedule picks the days and the hour, with presets for every day and weekdays. An event trigger fires on a new filing, an earnings press release, an earnings call transcript or an earnings filing, and email delivery is on by default. Its documentation, updated the same day, states that scheduled reports are available on the Standard and Pro plans and that agents on the Evaluation plan run manually.

Where it wins: the clearest self-serve path to an earnings-triggered report, priced in credits charged in line with the actual token usage of each run.

Where it falls short: no entitled broker research and no warehouse connection, so the report is the public record only. Its productivity figures (60 hours a week per analyst, 50 to 75 percent less document time) carry no cited source, so keep them out of a business case. No per-run company cap is published.

AlphaSense Workflow Agents

AlphaSense Workflow Agents, per its help center article updated April 9, 2026, manage end-to-end research workflows and produce reports, decks, memos, tables and slides with a single click. A user can run up to 10 agents at the same time. Scheduling is supported for Custom agents only, and not yet for the Organizational agents a team publishes under its "My Organization" tab.

Where it wins: the library the schedule runs over. AlphaSense holds licensed broker research and a 300,000-plus expert-transcript library (8,000-plus added monthly, per its site in August 2026), so a scheduled Custom agent asking what changed in sell-side tone on a name this week reads sources no first-party tool has. Generative Grid, capped at 400 documents by 12 prompts (help center, August 21, 2026), handles the bulk read after the quarter.

Where it falls short: no earnings-date or filing trigger is documented; the schedule is a clock. Internal content enters through SharePoint, Box, Google Drive, Egnyte and uploads, with no Snowflake or Databricks connector named on its Enterprise page. Pricing is quote-only; Vendr's $17,500 median contract is a third-party estimate from February 2026.

Daloopa

Daloopa is the model-update layer: source-linked historicals for 6,000-plus public companies with 14 years of history (company-stated, August 2026), every value hyperlinked to its location in the filing. Its MCP connectors put that data inside ChatGPT (December 2025), Perplexity (April 30, 2026), Microsoft 365 Copilot (June 25, 2026) and Gemini Enterprise (August 25, 2026). Microsoft's Copilot in Excel release of June 25, 2026 pairs the Daloopa connector with prebuilt skills named @model-update and @catalyst-calendar.

Where it wins: the filing step. Daloopa states that its data saves an average of two hours per ticker when updating models during earnings season, a company figure with no methodology published, and on the Walmart example below it is the tool that lands $187.9 billion in the model with a link to the page. Automating model updates covers the workflow in full.

Where it falls short: Daloopa publishes no sentence saying models update automatically when a company files; treat it as source-linked updates run from the tool or through a connector. It writes no brief, reads no transcript and holds no broker notes. The free plan is capped at three data sheets and paid tiers are quote-only.

Aiera

Aiera is the event and transcript feed under many of the tools above, including AllMind AI's live calls. It stated 45,000-plus events a year across 13,000-plus companies in October 2025, and its site shows 15,000-plus equities and 50,000-plus events in August 2026, with transcripts within one to three hours of an event (company-stated, October 2025). On June 3, 2026 it launched a consortium-backed content delivery platform with API and MCP connectors for Anthropic, OpenAI and Microsoft Azure.

Where it wins: the earliest trustworthy "transcript posted" event. Its March 19, 2026 vendor benchmark scored its human-reviewed transcripts 8.04 out of 10 against 5.74 for an unnamed provider, and a desk building its own pipeline gets the transcript through MCP the hour it is ready.

Where it falls short: Aiera is a pipe. It does not read broker notes or the house model, it emits transcripts and summaries, and it publishes no pricing; the only call to action on aiera.com is a demo request.

Hebbia Max

Hebbia introduced Max on July 30, 2026 as an agent that pulls from a firm's data, reasons through it with finance-specific agents and skills, and returns finished slides, a report or a financial model, rolling out to a small set of firms first. Its announcement lists slides, reports, Excel models and dashboards as outputs, and says Max works over email. A Snowflake integration shipped July 8, 2026.

Where it wins: the depth of the read on a defined document set. Matrix-style extraction, one KPI across every document in a folder with a citation per cell, is the private-equity and credit job Hebbia is known for, and a post-earnings grid across eight quarters of one company's filings is that job.

Where it falls short: Hebbia publishes no scheduled-run or earnings-trigger feature for Max; the page says firms can turn processes into agents that run on their own, without a trigger described. It brings almost no market data of its own, so the universe is whatever you load, and Max is not generally available as of August 2026.

Claude Managed Agents and OpenAI (the build-it-yourself tier)

Anthropic released ten finance agent templates on May 5, 2026, including an earnings reviewer and a meeting preparer. As Claude Managed Agents on the Claude Platform, the same template runs autonomously on a nightly schedule, with no published pricing. Claude for Financial Services connectors include Daloopa, Aiera, S&P Global, PitchBook, Snowflake and Databricks. OpenAI's ChatGPT for Excel (early March 2026) added financial data integrations, with premium providers requiring the user's own entitlement.

Where it wins: a team with two engineers and data contracts of its own can build a nightly run that pulls Aiera's transcript, Daloopa's numbers and the warehouse table at platform rates. The compliance guide for hedge funds using ChatGPT and Claude covers what the compliance file needs first.

Where it falls short: the trigger, the entitlement check, the audit log and the failure alert are yours to build and keep running. A connector's entitlement stops at that source, and neither vendor holds a licensed broker-research library, so the T+1 broker roll-up is out of reach without a separate contract.

What should an earnings-triggered AI research automation produce? A worked example on Walmart

An earnings-triggered AI research automation should produce two documents per print: a pre-read before the date and a review after it, each built from figures the reader can click through to. Walmart (NASDAQ: WMT) reported Q2 FY2027 results on August 20, 2026, one quarter after its Q1 release of May 21, 2026, so the pair tests both documents. The figures below are Walmart's own; nothing here is AllMind AI output.

The pre-read, T-3 (Monday, August 17, 2026). The question, as the covering analyst would save it: "Before Walmart reports on Thursday, restate last quarter's headline numbers, the guide for this quarter and the fiscal year, and the three things to check in the release." Walmart's Q1 FY2027 release of May 21, 2026 supplies every figure:

The review, T+0 (Thursday, August 20, 2026). The saved question: "Walmart has reported; lay the print against the Q2 guide, show every fiscal-year guidance change, and quote the language that changed." Walmart's Q2 FY2027 release answers it:

Walmart's release prints the guidance history as a three-column table (original February 19, 2026; as of May 21, 2026; as of August 20, 2026), the shape a guidance-change tracker across a coverage list should copy for every name.

What each tool does with those two documents: Quartr and Marvin Labs fire on the release and write from it. Daloopa lands $187.9 billion and $0.81 in the model, linked to the page. AlphaSense's scheduled agent reads the broker notes the next morning. AllMind AI's run adds the house model from the warehouse, joined to the release through the company object, plus the T+1 read of entitled broker notes. That join of external classes and the firm's own numbers on one ontology is what hedge funds and asset managers run these automations for.

How to automate earnings prep for every company in my coverage

Automating earnings prep for every company in a coverage list takes one specification per output, applied to the whole list, with the trigger, the sources, the template and the owner written down. The table below is the spec; copy it and fill the right-hand column per name. Each row is one automation, so a 60-name list needs five rows.

Three rules keep this from becoming five hundred emails.

1. Set a market-cap floor and a sector map on the automation itself, so the five rows route to the right analyst without forwarding.
2. Decide the exclusion list up front: dividend declarations, small insider buys, routine 8-K items.
3. Put the failure alert in the spec: a run that finds no transcript by T+0 plus four hours says so, in the same inbox.

The cost line is the reading. A five-page pre-read and a five-page review take about 20 minutes per print, so a 60-name list is roughly 20 analyst hours a quarter, whatever tool wrote them. The earnings-season preparation guide and the equity research workflow automation guide cover the manual version of each row.

What still needs an analyst after the automation fires?

The rating, the target and the narrative still need an analyst, and so do four situations that unattended runs handle badly.

When to stay with Quartr Automations: a global generalist desk covering names in 40 markets, where first-party material is the whole input and no broker contract exists. Quartr's roughly 15,000 to 16,000 company universe is wider than any first-party coverage AllMind AI publishes.

When Marvin Labs is the right answer: an individual analyst or a retail investor who wants the trigger on a monthly plan this afternoon and will read filings and transcripts only.

When a general assistant is the right answer: a team that already runs its own data contracts, has engineers to keep a nightly job alive, and wants to pay platform rates for a template it will change every month.

When to keep a person on the run regardless of tool:

• A late transcript, since the release-day brief will have blanks.
• A morning when 30 names in coverage report together, since caps are rarely published.
• A broker note on the aftermarket delay, which the T+1 roll-up will miss.
• Any print where the language moved more than the numbers: on Walmart's August 20 release the raised outlook is the headline, but the Flipkart timing shift and the tariff-refund reinvestment sit in the CFO's guidance paragraph.

Can ChatGPT or Claude run an earnings automation on their own?

Partly. Claude Managed Agents run a finance template on a nightly schedule, and Claude connects to Daloopa, Aiera and Snowflake, so a team with engineers can assemble a pre-read from those parts. Neither vendor supplies the earnings-calendar trigger with a lead time, an entitlement layer for broker research, or a per-run audit log a compliance officer can read.

The practical split in 2026: use ChatGPT or Claude to draft and restructure what a platform run produced, and use a platform (AllMind AI, Quartr, Marvin Labs or AlphaSense, depending on sources) to own the trigger and the log. How ChatGPT, Claude and Perplexity compare with institutional research platforms sets that boundary out in full.

Frequently Asked Questions

What are the best tools for scheduled research automations?

AllMind AI is the pick for an institutional desk that wants a pre-read before each print and a review after it, across a coverage list, with broker research and the firm's own model in the same run. Quartr Automations, launched August 24, 2026, fits teams that live in company-published material. Marvin Labs fits one analyst on a monthly plan, AlphaSense schedules Custom Workflow Agents over its licensed library, and Claude Managed Agents fit a team with engineers who will own the trigger and the audit trail themselves.

What are the best Quartr Automations alternatives?

The closest alternatives are AllMind AI for a coverage-wide brief that reads entitled broker research and internal data, and Marvin Labs for filing and transcript triggers on a self-serve plan (July 27, 2026). Aiera supplies the raw event feed, with transcripts one to three hours after a call. Quartr keeps two advantages: a trigger that fires the moment a company publishes and coverage of roughly 15,000 to 16,000 companies across 60-plus markets. None of the alternatives matches that market breadth on first-party documents.

Is there an AI agent that runs automatically when a company reports earnings?

Yes, and by August 2026 four vendors document one. AllMind AI hangs the trigger off the earnings calendar with a lead time you choose, so the pre-read runs days before the print and the review runs when the release, the transcript and the filing land. Quartr fires the moment a company publishes, Marvin Labs fires on a new press release, transcript or earnings filing, and Claude Managed Agents run a template on a nightly schedule. The tools differ on what the agent can read once it wakes up.

How many companies can one earnings automation cover?

AllMind AI states on its comparison pages that one automation tracks event triggers across up to 200 companies, with the run delivered as Word or PDF. AlphaSense caps concurrent Workflow Agents at 10 per user and schedules only Custom agents. Quartr, Marvin Labs and Aiera publish no per-automation company cap, and a do-it-yourself Claude or OpenAI pipeline covers whatever your code loops over. Ask every vendor what happens on a day when 40 names in your coverage report at once.

What does an earnings-triggered AI research automation cost?

Marvin Labs publishes the clearest self-serve basis: scheduled and event-triggered reports on its Standard and Pro plans, billed in credits against actual token usage, with the free Evaluation plan manual only. Quartr Automations requires Quartr Pro, which is quote-only, and AlphaSense, Hebbia, Aiera and AllMind AI are all enterprise quotes. Hudson Labs lists a Core plan at 99 dollars a month and reserves unlimited agentic automations for its quote-only Institutional plan. Budget the analyst hour that reads the output, since every one of these still needs a reader.

AllMind AI is the AI-native research platform for institutional equity teams. If you want proof on your own work, send us the workflow you want tested.