Superhuman Added a Knowledge Base. Here's What It Actually Changes.

Superhuman's March 2026 launch of Knowledge Base (BETA) makes this comparison worth revisiting. For the first time, Superhuman users on Business and Enterprise plans can upload files and connect links — compliance docs, pricing sheets, help center pages — so AI drafts pull from real business context.

That's a meaningful move. It also sharpens the line between what Superhuman is building and what we're building with Inbox SuperPilot.

The short version: Superhuman optimizes email workflow. We optimize email content — grounded in your actual docs, CRM data, and cited sources.

Those sound similar. They're not the same product strategy.

The job-to-be-done split

Superhuman is becoming a broader productivity suite — Mail + Grammarly + Coda + Go — with AI agents and cross-app acceleration layered on top. Their new Agent Store includes partners like Box, Fireflies, Common Room, and Gamma. That's real breadth, and for teams whose bottleneck is inbox triage and context-switching, it's a compelling answer.

We built Inbox SuperPilot for a different job. If you sell, support, or advise from a mix of pricing sheets, product docs, help center content, CRM notes, and internal policies — and you need every outbound email to be accurate and verifiable — that's the problem we solve. Nothing more, nothing less.

If you need cross-app agents and a unified productivity suite, we're not that product. We're not trying to be.

Knowledge Base is context. Citations are proof.

Here's the core distinction as of their March 2026 launch: Superhuman's Knowledge Base makes drafts more informed. But their documentation doesn't emphasize inline source citations in the draft itself — you still have to open the original doc to verify details. That keeps review in "fact-checking" mode instead of "verify in seconds" mode.

We took a different approach. Every Inbox SuperPilot draft includes clickable source citations linking back to the specific document, page, or article the claim came from. You're not asking "is this probably right?" You're clicking through and confirming it matches the source.

That changes the review workflow entirely. For teams handling customer pricing, implementation commitments, or policy details, that's the difference between trusting a draft and trusting a process.

Manual uploads vs. automatic retrieval

Superhuman's KB requires manual file uploads and link connections. That gives teams control over exactly what's in the knowledge base — and for some organizations, that curated approach feels safer.

We connect to 20+ knowledge sources — Google Drive, Notion, Confluence, Dropbox, websites, help centers — through an automatic RAG pipeline. When your team updates a pricing doc in Drive or a policy in Notion, the next draft reflects those changes without anyone re-uploading anything.

I'll be honest about the tradeoff: automatic retrieval introduces its own risks. A RAG system can surface the wrong document, pull from stale cached content, or retrieve something outside the intended scope. We mitigate this with hybrid semantic + keyword retrieval and source citations that let you catch errors before sending — but it's not a zero-risk system. No AI drafting system is.

The practical question is whether your team will maintain a manually curated KB over time. In our experience with teams we've worked with, manual knowledge bases often start strong and decay fast. Automatic retrieval trades some curation control for freshness and breadth.

CRM context belongs inside the draft

Both products integrate with HubSpot, Salesforce, and Pipedrive. The difference is what happens with that data.

Superhuman's CRM integrations are read-only — you can see deal info alongside the email. Inbox SuperPilot pulls CRM context directly into the draft itself.

Here's what that looks like concretely: when you reply to a prospect in Stage 3 of a deal, the draft references their specific pricing tier, notes from the last call, and prior commitments your team made. You're not staring at a sidebar doing mental copy-paste — your reply already reflects the deal stage, last conversation, and promises you've made. It's not a generic follow-up with deal data visible in a sidebar. It's a draft that already knows the relationship.

For revenue teams, that's a meaningful difference between "has CRM integration" and "uses CRM data to write better emails."

Gmail vs. dedicated app

Superhuman replaces Gmail with its own client. We work inside Gmail as a Chrome extension.

For most Gmail-native teams, replacing Gmail is a harder sell than adding an extension. That's our bet. It affects rollout friction, training cost, and day-to-day adoption in ways that matter more than interface preference.

Some teams genuinely prefer Superhuman's opinionated, dedicated environment. That's a legitimate choice — not everyone wants to stay in Gmail.

Pricing and access

Superhuman starts at $30/month for Mail. Their free tier exists but doesn't include Mail — only Go, Grammarly, and Coda. Knowledge Base requires Business ($33/month annual) or Enterprise.

Inbox SuperPilot is free forever — 50 drafts/month, all features, no credit card. Pro is $25/month.

If you want to test KB-grounded drafting inside your existing Gmail workflow with zero friction, we designed for that.

What happens right before you hit send

Every AI email product has to answer a trust question: what happens between draft and send?

Our answer is explicit. We create drafts for your review — we never auto-send. Quality Guard checks whether every question in the thread was answered, flags potential PII, and catches missing attachments. Combined with source citations, the goal is making the human reviewer faster and more confident, not removing them from the loop.

The honest recommendation

Choose Superhuman if your bottleneck is inbox overload, workflow speed, cross-app productivity, and you want a premium all-in-one suite. It's a strong product getting stronger.

Choose Inbox SuperPilot if your bottleneck is drafting accurate replies from docs and CRM, you need to verify claims before sending, and your team lives in Gmail. That's the narrower, sharper problem we built for.

For teams where a wrong email means a wrong price quoted, a wrong policy cited, or a wrong commitment made — "faster email" and "accurate email" aren't the same thing.

If you're fact-checking AI drafts before every send, that's the problem we built for. Try it free inside Gmail →

Further Reading & References

From the Inbox SuperPilot Blog

• What We Learned Building a Citation Engine for Email AI
• Gmail Gemini vs. Knowledge-Grounded AI: An Honest Comparison
• Why Generic AI Fails in Customer Support Email Workflows
• The Hallucination Problem

External References

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks — Lewis et al., 2020. Why grounding generation in retrieved documents improves factual accuracy.
• RAGAS: Automated Evaluation of Retrieval Augmented Generation — Es et al., 2023. Framework for evaluating retrieval quality and answer faithfulness.
• Factuality Challenges in the Era of Large Language Models — Augenstein et al., 2023. Survey of hallucination types and mitigation strategies in LLMs.