Amendant turns your scattered documents, codebases and tribal knowledge into a single, self-correcting source of truth. Then the same brain answers customers and staff across chat, voice, email, WhatsApp and four other channels — in your tone, with citations, on every device.
Most teams don't have a knowledge problem. They have a knowledge debt problem: dozens of overlapping documents, written by different people, in different years, that contradict each other in ways no one has noticed yet.
Send us a sample of your docs. We'll send back a map of what's covered, what's missing, and what's silently fighting itself.
Not a list of features. The three changes a non-technical manager can actually point to and say: that's different now.
Every fact in your business gets one canonical version. When two documents disagree, Amendant tells you before a customer notices. When something goes stale, you get a Slack digest.
One knowledge base, seven channels. Web chat, phone, email, SMS, WhatsApp, Slack, Telegram — same brain, consistent answers, your brand's voice across all of them.
Confidence scoring decides which messages send themselves and which sit as drafts in your inbox. Your team stays in the loop on anything that matters — and out of the loop on anything that doesn't.
Amendant maintains two separate knowledge bases for every tenant — one for the world, one for your operators. The same intelligence layer, fed by different sources, accessible to different people.
Product knowledge, policies, codebase docs, public FAQs. Powers the chat widget, phone line, email inbox and every other channel.
Workflow configs, connector status, KB health reports, internal SOPs. Powers the Platform Agent — a chat interface where your admin asks "what's in my review queue?" or "design an interview scenario for new-hire onboarding".
Amendant's defining mechanism. Every time you add a document, we run a four-step check against everything already in your knowledge base. Contradictions ping your inbox before they ping your customers.
Most AI assistants are confident. Amendant is correct — and when it isn't sure, it says so, books a callback, and tells you what's missing from your knowledge base.
Channel-specific prompts adapt the tone — voice gets three sentences, email gets a formal one-shot reply, chat gets structure and images. The knowledge stays identical.
No data scientists, no six-month migration. Most teams are answering real customer questions within their first week.
Drag in PDFs, Word docs, slides, spreadsheets. Connect Notion, Slack channels, Guru, Freshdesk, your codebase. Scanned pages, embedded tables and images are extracted automatically.
Amendant clusters your content by topic, flags contradictions, and asks a human to pick a side, merge, or exclude. You stay in control of what counts as "official".
Brand the chat widget, pick the channels, choose escalation rules and authorization mode. Set monthly quotas and a confidence threshold for auto-send.
One script on your website. One number for voice. One inbox connected. Amendant starts answering and keeps learning what you knew all along.
Every plan includes all seven channels, the full KB pipeline, dedicated tenant isolation, and the admin console. You pay for traffic, not for headcount.
And the value of a fast, accurate, in-brand one is higher. Amendant earns its keep first in industries where expertise is the moat.
Especially teams maintaining legacy applications. Amendant absorbs method-level signatures and product docs so support reps and coding agents both query the same source of truth.
When client data lives in Guru, project notes in Notion and contacts in HubSpot — Amendant becomes the orchestration layer that unifies them without forcing a migration.
Where exact wording is non-negotiable. Document Immutability locks contractual, regulatory and policy text so Amendant ingests it verbatim — never paraphrased.
Security is architecture, not a checklist. Hard code-level constraints sit between the LLM and your data — not prompts asking it nicely.
Only the customer-support and booking agents are reachable from chat. Even if the LLM is tricked, the router refuses to call admin or developer tools.
Twenty-five attacks across eight categories — instruction overrides, tool abuse, tenant escape, confused deputy — run continuously. Failures break the build.
Mark a document confidential and it's filtered from unauthenticated sessions. Authenticated users get a session-validated proxy URL — no public S3 links.
Deploy in AWS or Azure, in the region of your choice. Your DynamoDB, S3 bucket and vector namespace are physically yours — not shared with another tenant.
If yours isn't here, ask us directly — we'd rather answer it now than have it sitting in a procurement spreadsheet.
No — that's not the goal, and the design reflects it. By default Amendant runs in human-in-the-loop mode: high-confidence answers send themselves, anything below your threshold sits in your team's inbox or help desk as a perfect draft.
Teams use the time we give back to handle the conversations that actually need a human — escalations, edge cases, relationship work. Most clients keep the same headcount and double the volume they can handle.
Every tenant gets physically isolated infrastructure: a dedicated DynamoDB table, a dedicated S3 bucket, a dedicated vector namespace. We don't multiplex you onto a shared database.
You pick the cloud (AWS or Azure) and the region. For regulated industries we deploy single-tenant infrastructure with custom data-residency guarantees.
The full audit pipeline runs in under an hour on a typical mid-market document set. Once you've reviewed the conflicts, a basic deployment — chat widget on your site, email channel connected, brand colours applied — takes another day or two. Most clients are live within their first week.
Yes — that's exactly what the integration layer is for. We have native adapters for HubSpot, Salesforce, Microsoft Dynamics and GoHighLevel. For other CRMs we provide a webhook endpoint that works with Zapier, Make, or your own glue. We don't ask you to migrate anything.
First, it usually catches itself. Amendant's confidence scoring distinguishes "I know this" from "I'm guessing" — and below the threshold it either escalates to a human or asks a clarifying question rather than confabulating.
If a wrong answer does ship, the feedback loop catches it. Thumbs-down feedback adjusts the underlying fact's score, the KB Analyst Agent surfaces it in the weekly digest, and your team can retire or override the offending fact in seconds.
Generic chatbots are language models — they answer from what they were trained on. Amendant is an expertise platform: it answers from your documents, with citations, with conflict detection, with verbatim guarantees on regulated text, and with security architecture built for tenant isolation.
Concretely: a 2-hop knowledge graph, cross-encoder reranking, fact signatures that persist across rebuilds, the AUDN loop, code-level agent lockdown. None of which a generic chatbot offers.
Everything important happens in a web dashboard. Knowledge managers upload documents and review conflicts in a UI. Admins configure channels, escalation rules and quotas with checkboxes. The only "code" most teams ever touch is a one-line embed script for their website — and we'll paste it in for you.
Yes — that's the Platform Agent. Every tenant gets a second knowledge base (the platform KB) and a chat interface where admins can ask things like "what's in my review queue?", "design an interview scenario for new-hire onboarding", or "create a workflow that triggers on the word 'enterprise'." No configuration UI required.
Send us a sample of your documents. In 60 minutes we'll send back a patchwork report — the contradictions, gaps and tribal-knowledge spots in your own KB.
