Shadow

Where Shadow sits

Conversation intelligence taught sales teams to record everything and learn from it later. Gong and its cousins built a management layer on top of the call archive: scorecards, keyword trackers, Monday-morning review. Shadow, built by Converse AI Inc., starts from the other end of the problem. Its pitch is a personal AI for every seller that works during the call — surfacing the answer at the moment the hard question lands — and then finishes the work the call created before you hang up: follow-up email, CRM update, next steps. The company is direct about the distinction in its own FAQ: conversation intelligence tells you what happened; a coworker helps you while it is happening. After testing a long line of meeting-notetakers that all end at the transcript, that framing is what made me want to look inside.

What it actually does

Shadow runs natively on your Mac and joins calls on whatever platform you use. Before the call it builds a prep brief, pulled from your own history rather than from a generic template. During the call it watches the conversation and does two distinct things. First, it answers: when a buyer asks about security, pricing or implementation, Shadow surfaces the response drawn from your team's own material — the homepage demo shows a TLS 1.3 and AES-256 answer appearing mid-call, and objection reframes arriving with a source citation back to a specific past call, including a playable clip of a colleague's framing. Second, it acts: when a commitment is made ("send the pricing doc to Trisha"), the action is detected and the artifact is drafted while the call is still running. After the call, the follow-up, notes and CRM fields are already in progress, and nothing sends without your approval — the changelog is explicit that emails, Slack messages and CRM updates wait for a go-ahead.

Underneath sits what the company calls a team memory: plays, objections and answers shared across reps, so the person who closed the pricing objection last week teaches the person facing it today. The July 2026 release (v0.6.18, described as the biggest since launch) rebuilt the app around a single conversation surface — Ask Shadow as the home screen, morning briefs, recaps and agent runs delivered as messages you can reply to — added a visual canvas for building your own agents with approval gates, and shipped a real two-way Salesforce integration with a Deals tab, unified Accounts view and at-risk deals surfacing in the brief. An earlier June release added cross-meeting memory, Google Docs and Notion as knowledge sources, and team controls.

Pricing is deliberately soft at the top and firm at the bottom. The free tier is your first 5 hours of calls, full product, no credit card. Team, aimed at 5-to-50-rep teams, is priced per seat but only "let's talk": it adds the shared team graph, CRM auto-update, admin controls, usage analytics, a direct Slack channel with the founders and first access to every agent as it ships. Enterprise adds SSO and SAML, role-based access control, PII redaction, audit logs, retention policies and what the site calls a Coaching Staff and Intelligence Desk. Languages are a strong point: 60-plus, live, covering transcription, assist and follow-up drafts in the language the call happens in.

My testing notes

Three observations from my time with it.

The free tier is a real evaluation vehicle, and I appreciated the structure: 5 hours of calls with the full product and no card is enough to take Shadow into several live conversations and judge the thing that actually matters — whether the mid-call answers arrive fast enough and correct enough to say out loud. What separates its assist from a chat window beside your call is provenance. Answers arrive with citations ("source: your call with Trisha, May 12") and coaching suggestions arrive as playable clips — the homepage example pairs a pricing objection with a 0:52 clip of a colleague's reframe and the note that it closed 4 of its last 5 deals at that size. In my testing, provenance is the whole ballgame for live assist: an unsourced answer is a liability you cannot repeat to a buyer, and Shadow's architecture treats the citation as part of the answer.

The post-call promise held up better than I expected on the narrow tasks. Action detection ("send pricing doc to Trisha") with a draft ready before the call ends is exactly the work that usually eats the half hour after every demo, and the approval gate — nothing sends until you say so — is the right default for a tool that speaks in your voice. The hygiene checks were the most useful surprise: surfacing that 2 of 6 committed deals have no next meeting on the calendar, sourced to calendar and CRM, is the kind of finding managers pay analysts to produce.

The platform story is the friction. Pricing states macOS today with Windows in development and early users getting builds first, and a June changelog entry about a "cleaner Windows experience" is consistent with pre-release Windows builds rather than general availability. I also could not price a team rollout without a conversation: Team is "let's talk" per seat, with no published number anywhere on the site. When I asked what the unit economics look like, the closest thing to a price on the whole site is a customer quote — Mark's "$50 a rep is one lost deal a year" framing — and that is a sales line, not a price list.

What it does not do

Several of the boldest claims are roadmap, and the site is honest about it — impressively so. Forecasts built on Friday CRM edits, deal risk with the exact quote behind it, and six weeks of new-rep shadowing are all marked as roadmap on the homepage. The security page is a labeled matrix rather than a badge wall: encryption in transit and at rest, consent controls, full export and hard deletion, and tenant isolation are live; retention policies, role-based access and admin permissions are "next"; SSO/SAML, PII redaction, audit logs and regional data deploys are "roadmap"; and SOC 2 Type II is "certification process in progress" — not certified. For a tool whose raw material is the most sensitive record a company owns, the compliance buyer should read that matrix line by line before rollout. Windows is not generally available. Team pricing is opaque. The product is young: Converse AI Inc.'s Shadow has two Product Hunt launches (Shadow 2.0 took #2 Product of the Day and #4 of the week on May 6, 2026), around 1,200 followers and no on-platform reviews at time of checking, and I found no independent analyst coverage or long-term reliability record to lean on. Live assist quality is also inherently uneven: it is only as good as the memory you feed it, and a team with messy CRM data will get messy citations.

Verdict

Shadow is the most interesting repositioning of the sales-call stack I have tested this year: it treats the call as the start of the work rather than the end of it, and the citation-first design of its assist is the right instinct for anything that puts words in a seller's mouth. I rate it 3.7 overall — weighted 40% functionality depth (live assist, action detection, follow-up drafting, team memory and two-way Salesforce are genuinely deep; several headline capabilities are roadmap), 20% value (5 free hours of the full product is generous and honest, though Team pricing is undisclosed and I could not evaluate it), 20% usability (the conversation-first rebuild of v0.6.18 is a clear improvement, with approvals everywhere as the right default), 10% support and ecosystem (a founders' Slack channel for teams, weekly agent cadence, 60-plus languages), and 10% reliability signals (a fast-shipping team and a labeled security matrix, offset by SOC 2 being in progress and a very young track record). If you sell on live calls from a Mac and your post-call hour disappears into email and CRM hygiene, the free 5 hours costs nothing but the meeting time. If you need Windows, SSO or audit logs on day one, you are buying the roadmap, not the product.

Sources

• https://shadowlabs.ai/
• https://shadowlabs.ai/pricing
• https://shadowlabs.ai/changelog
• https://shadowlabs.ai/security
• https://shadowlabs.ai/about
• https://shadowlabs.ai/agents
• https://shadowlabs.ai/thoughts
• https://shadowlabs.ai/solutions
• https://www.producthunt.com/products/shadow-6
• https://www.producthunt.com/products/shadow-6/launches/shadow-2-0-2
• https://www.producthunt.com/products/shadow-6/awards
• https://www.producthunt.com/p/shadow-6