inference

Ergo — subprocessors

Last updated: 6 August 2026

Ergo is an AI marketing advisor that works inside a Slack channel, built and run by TCHQ Ltd, trading as Inference. This page lists the companies involved in running it, what each one can hold, and where.

It accompanies the Ergo for Slack privacy notice, which explains what happens to your conversation. This page sets out the supplier detail behind it, and is also how clients are notified when that list changes.

#Subprocessors

These are the companies that can hold or process client data as part of running Ergo.

In summary: the database holding your strategic context and conversation summaries, and the storage holding the documents Ergo publishes, are both in the UK. A search index over that same context is held in the United States, as are the companies that process a conversation in order to answer it, that carry out web searches, and that connect Ergo to any systems you have linked. Usage statistics are processed in the European Union. Each is set out below.

Subprocessors engaged by TCHQ Ltd for the Ergo service.
CompanyWhat it does for ErgoWhat it can holdWhere
Vercel Runs the Ergo application, stores the documents Ergo publishes, and provides the AI Gateway that routes messages to models. Conversation content while a request is being handled; published documents at rest. Vercel Inc. is a United States company. The application and the document storage both run in London (lhr1).
Neon The Postgres database behind Ergo. Strategic context, conversation summaries, links back to Slack threads, escalation records, client identifiers. Neon Inc. is a United States company. The database is located in London (AWS eu-west-2), so stored data remains in the UK.
Supermemory A search index over your stored context, so Ergo can find relevant history. An index over your strategic context and conversation summaries. Separated per client. Supermemory, United States.
Model providers Generate Ergo’s answers, and categorise conversations for the usage statistics. Reached through the Vercel AI Gateway — see below. The conversation thread and any files in it, for the duration of the request. Not retained, not used for training. Varies by model. See below.
Exa Web search, when Ergo researches something current. A search query derived from your question. Not the conversation. United States.
Firecrawl Renders web pages Ergo cannot read on its own, and returns the text. The web address only. Not the conversation. United States.
Composio Connects Ergo to systems you choose to connect, and passes data between them and Ergo. The access credentials for those connections, and data passing between the connected system and Ergo. Keyed to your client identifier and unreachable from another client’s session. Composio, United States.
PostHog Records the usage statistics described in the privacy notice. Counts and categories only. No message content and no individual user identifiers — records identify the client organisation, not the people in the channel. European Union.

#AI models

Ergo doesn’t contract with model providers directly. It reaches them through the Vercel AI Gateway, on terms that mean your conversation is used to generate the answer, isn’t kept afterwards, and isn’t used to train anything.

Ergo’s model selection changes as improved models are released, sometimes month to month. Naming them here would require a subprocessor notice on each change, which would reduce the value of those notices. What is disclosed is therefore the route: the gateway is the contracted supplier, and models are drawn from those available within it.

The models in use at any time are available on request. If your organisation requires individual approval of model providers before use, this can be addressed in your contract.

#Slack

Slack is not listed above, and its omission is deliberate. Ergo operates inside your Slack workspace, under your agreement with Slack. Slack is your supplier rather than ours: where your Slack data is held, and how long Slack retains it, is determined by your own arrangement with Slack and not by us.

What we do is read the threads Ergo is mentioned in, as described in the privacy notice.

#Connected tools

Composio, listed above, is the connector infrastructure and is part of the standard service. The individual systems you choose to connect are not listed above, because they are not part of the service everyone receives but an option you have enabled. What each can access is determined by the permissions you granted when connecting it.

If you connect a system, personal data in that system can reach Ergo. If you connect an inbox or a CRM, that includes personal data about your customers and correspondents.

#Other services

These are not subprocessors of personal data. They are listed for completeness because they are reached while the Service is in use.

esm.sh. When you open an editable diagram Ergo has produced, your browser loads the diagram software from this public code distribution network. Your browser’s IP address reaches them. Nothing else does.

#How changes are notified

This page is the notification mechanism, and the data processing agreement points at it. When a company is added, removed or swapped, this page is updated, the date at the top moves, and the change is recorded below.

Clients receive advance notice of additions, with time to object, on the terms set out in the data processing agreement. If you would prefer direct notification rather than checking this page, contact us and we will add you to the notification list.

#Change history

DateChange
6 August 2026 Added Supermemory, Composio, PostHog and esm.sh. All four were in use when this page was first published and were omitted in error, so this corrects the list rather than reporting a change of supplier.
6 August 2026 First published.

#Contact

privacy@thisisinference.com
TCHQ Ltd, 24 Hove Park Road, Hove, England, BN3 6LJ