Your Salesforce has years of configuration on top — sharing rules, FLS, Apex triggers, custom objects. We connect AI agents and workflows against your instance respecting every last permission, without Einstein and without replacing anything.
The AE opens the Account, jumps to Opportunities, checks the latest Activity, drops into the open Task, goes back to the Account to look at the contracts custom object and only then starts preparing the call. Multiply by forty accounts. The SDR triages new Cases by hand. Marketing asks for proposals the team writes from scratch because the Pricebook is buried two tabs away.
We build the connector against your Salesforce org via a Connected App and OAuth, with an integration user that respects Field-Level Security, record-level sharing and every profile permission in your model. The agent never sees a field the user’s role couldn’t see. It reads and writes to Salesforce with idempotent operations, audit trail through FeedItems and deploys via change set or package — the same enterprise flow you already use for Apex.
Reads the Account, Opportunities, Activities, Tasks and relevant custom objects and summarises the state of the account in one actionable paragraph. Before every call, the AE walks in with context instead of rebuilding it by hand.
Takes the Opportunity, reads OpportunityLineItems, crosses with your active Pricebook and generates the commercial proposal in your template. The AE reviews, adjusts the discount and signs — the document is already built.
Crosses usage signals, open Cases, NPS and commercial activity to score churn risk per account. Proposes the next action to the CSM (health call, discount, escalation) and drops it as an assigned Task.
Every new Case is classified by type, urgency and product, routed to the right queue and prioritised. Repeat cases link to the relevant KnowledgeArticle so the human agent starts with the answer in hand.
Before every meeting, reads the Account, Contacts, latest engagements and the customer’s public web. Generates a 5-bullet brief with who the account is, what happened last time and what to propose now.
We create a dedicated Connected App in your org with the minimum OAuth scopes (api, refresh_token, optionally offline_access). No shared generic users, no passwords in code. Token revocable from Salesforce at any moment.
The integration user has dedicated profile and permission sets, mapped to the role of the user the agent acts on behalf of. The agent never reads a field restricted by FLS or a record outside its sharing — Salesforce stays the source of truth on who sees what.
Mass flows (overnight enrichment, scoring) go via Bulk API 2.0 to avoid burning API calls. Real-time flows (Case triage, Opportunity stage change) go via PushTopic or Change Data Capture depending on the object. Each one where it fits best.
We don’t invade your Apex code. If a use case requires logic that already lives in a trigger or a class, we call an Apex REST endpoint your team controls. The line between what the agent does and what your Apex does stays explicit.
We deploy to sandbox (developer or partial copy depending on the case), validate against real data and promote to production via change set or unmanaged package. The release process is the one your team already uses — we don’t introduce a parallel channel.
Yes, no exceptions. The integration user the agent runs as has dedicated profile and permission sets, aligned with the role of the human user the case is about. If an AE can’t see the Margin field on their Opportunity, the agent acting on their behalf can’t see it either. We audit permissions before going live and document them so your Salesforce team can review them whenever they want.
No. We work against Salesforce via standard APIs (REST, Bulk, Streaming) and bring the intelligence from the model that best fits the case. Einstein is a valid option if you already have it and it makes sense for your flow, but it’s not a requirement and we don’t sell it as ours. If at some point Einstein solves the problem better than an external model, we say so.
Not by default. The connector lives outside your Apex code and talks to it via REST endpoints when the case needs it. If the ideal flow goes through an existing trigger, we call the public method your team exposes — we don’t change classes or triggers without you. Any Apex modification follows the same process as any other deploy from your team.
We always start in sandbox (developer or partial copy depending on the data volume we need to validate). We build, test against copied real data and measure for an agreed period. Promotion to production happens via change set or unmanaged package, depending on what your team uses. Every release is documented with which objects it touches, which permissions it requires and how rollback works.
We design for privacy from the start, human control, traceability, usage limits, permissioning and documentation. For sensitive processes, we help assess risk and applicable obligations under GDPR and the EU AI Act.
Every engagement is led personally by one of the partners. If there's a fit, you get a personal first read of your case within one business day — not a canned demo.
