Aimdoc
AI sales agent that works the inbound side rather than the outbound one. The agent ingests a company's website, documents and FAQs, then converses with visitors, qualifies them against criteria the customer defines, books meetings through a calendar integration, and escalates to a human with a Slack notification when a person is needed. Agent voice and persona are configurable. A second product line extends the same agent into the application for trial onboarding and activation.
The company positions itself against inbound conversion incumbents, publishes a documentation site, case studies and an ROI calculator, and exposes both an MCP server and a public agent gateway endpoint so other AI agents can converse with its sales agent directly.
Capability Axes
Capability grades
17 of 17 axes rated · 8 graded A or B
The first A on this axis in the index, and it is earned by construction: the product is a conversational agent and there is no non AI remainder. Remove the model and nothing is left to sell, because the entire value is an agent that ingests a company's site and documents in minutes and then holds qualifying conversations with strangers.
What the AI does is documented rather than gestured at: ingestion of website content and uploaded material, question answering grounded in it, qualification against customer defined criteria, meeting booking, escalation triggers, session transcripts, intent capture and lead scoring.
The autonomy boundary is described in useful terms. The agent runs unsupervised against live visitors, but within customer defined qualification criteria, and the handover model is explicit: the agent recognises when a person is needed, notifies the team in Slack, and passes the conversation to a human. Routing to the right representative is rule based. What is not documented is containment. Nothing states what a misbehaving agent can commit to, what it is prevented from saying about price or product, or what stops a bad session once it is live in front of buyers.
Grounding is described, in that the agent is trained on the customer's website, documents and FAQs and suggests answers for gaps found in real sessions, which is more than nothing. The stack is not. No model or provider is named anywhere public, no version or update policy is stated, and the lead scoring logic that decides which conversations reach a representative is unexplained. A buyer cannot tell what is producing the words their prospects read.
A case study library and an ROI calculator exist, and the testimonial volume is high, but what is published is enthusiasm rather than measurement. The claims located run to increases in leads and bookings exceeding expectations and the tool paying for itself, with no conversion rate, no lift against a stated baseline, no sample and no period.
For a product whose whole promise is converting a known quantity of existing website traffic, a measured conversion rate against a holdout is the obvious evidence, and it is the one thing not offered.
Mostly inbound, which narrows the regulated surface: the visitor initiates and the agent responds. The product does extend to personalised outreach agents reachable by public link that customers place in emails, and nothing published addresses the sending obligations that come with that, or how consent captured in a chat session propagates into a customer's marketing suppression. No public enforcement action located.
Stronger than most vendors of this size and short of the top band on the question that matters here. Published: standard data processing agreement availability, data subject access request support, international transfer mechanisms, data minimisation and deletion of customer data on request. The gap is lawful basis for the people this product exists to identify.
The company markets anonymous visitor identification and account identification, and its published privacy documentation is a third party generated policy that says nothing about the basis on which an unidentified website visitor is resolved to a person or an account.
The knowledge side is unambiguous, since the agent is grounded in the customer's own website and uploaded documents. The identity side is not. Anonymous visitor identification and account identification are marketed capabilities, and no public material states where the identity resolution data comes from, whether it is licensed, contributed or reverse looked up from network signals, or who stands behind it. A buyer inherits that question along with the product.
Low exposure. The agent runs on the customer's own web property and inside the customer's own application, and reaches CRM, calendar and messaging systems through their documented interfaces. No automation of a third party network that prohibits it. No conformance position of its own is published, so the grade rests on architecture rather than a stated commitment.
The security program covers deletion on request, audit logging and least privilege access, which is the operational half. The cross client half is unanswered. Nothing published states whether visitor conversations, the questions prospects ask or the knowledge bases customers upload inform models or improvements that serve other customers, and no opt out or contractual exclusion is documented. For a product that describes its AI as continuously learning from successful interactions, that is precisely the question a buyer should ask before pointing it at their pipeline.
Nothing published states whether a website visitor is told at first contact that they are talking to software. The product ships configurable agent voice and persona across formal, casual and professional registers, and the customer testimonials the company features praise how human the agent feels, which is the marketing frame rather than a disclosure position.
Under the European Commission's final Article 50 guidance, in force since 2 August 2026, a system interacting directly with a person must disclose its artificial nature at first contact in the interface, and the obligation falls principally on the provider rather than the deploying customer.
Specific rather than logo listed: named object level behaviour in the major CRMs, creating and updating contacts, companies and deals in one and leads, contacts and opportunities in the other, plus calendar booking, Slack escalation notifications, webhooks and a REST API, with a public documentation site and an integrations page. Notably it also publishes an MCP server and a public agent gateway endpoint, so an external agent can query the sales agent directly. Depth thins on sync direction, field mapping and failure behaviour, which are not documented before purchase.
Multi tenant cloud described as running on the edge for response speed, which is a performance statement rather than a residency one. No hosting region, no European residency option, and no statement of where conversation transcripts and uploaded knowledge bases are stored. The security page refers to international transfer mechanisms where applicable, which acknowledges the question without answering it.
Genuinely good for a company this size. SOC 2 Type II certification is claimed with a dedicated trust portal, and the security page documents practice beyond the badge: background checks and security training, least privilege and role based authorisation, change management with peer review, vulnerability management, business continuity and incident response, TLS 1.2 and above in transit with encryption at rest, segregated production, secrets management, and backups with restore testing. Short of the top band on two points: the trust centre requires requesting access rather than serving documents, and no audit date or report recency is visible to an outsider.
An entry price is published, at 999 dollars a month for the first tier, alongside a 14 day free trial and an ROI calculator, with the two upper tiers quote only. One mechanical detail is worth a buyer's attention and is unusual enough to record: the pricing page renders its plans client side, so the served page reads as loading and the only price in the fetched document sits in the meta description. A vendor that publishes an MCP server and an agent gateway so machines can talk to its sales agent has left its own prices unreadable to a machine fetching the page.
Deletion of customer data on request is committed to on the security page, which covers one half of offboarding. The other half is missing: no documented export path for conversation transcripts, captured leads, intent data or the knowledge base a customer has built, no post termination data rights, no deletion timeline, and terms of service delivered through a third party policy generator rather than as negotiated commercial terms.
Sending is peripheral to a product that mainly answers rather than initiates, but it is not absent: personalised agent links are distributed by email and meeting confirmations flow to prospects. No sending infrastructure, authentication guidance or reputation practice is documented. Deliverability is neither claimed nor addressed.
The intended buyer is clear and the packaging corroborates it: business to business software companies with meaningful existing website traffic and a trial motion, positioned explicitly against the inbound conversion incumbents, with an entry price that rules out the smallest teams. A case study library and customer page back the claim. Geographic and language coverage is the gap, with no stated regional support, no language list for agent conversations, and no indication of where implementation help comes from.
Compared With
Editorial comparisons are published only where the index assesses two vendors as direct competitors for the same buyer. Each carries a verdict, the buyer conditions that favor each vendor, and a graded side by side.
Pricing
What this vendor charges, what it commits to in writing, and where the bill can move. Figures the vendor publishes itself are labeled Vendor Published. Figures labeled Estimated come from other sources and the vendor has not confirmed them.
- ›The price is $999 a month for the entry offering, and the only place it exists is in the page's description and its structured data. Open the page in a browser and you see a trial button and no price at all.
- ›That means an AI assistant can tell you what this costs quite confidently while a human visitor cannot find out. It is the usual metadata problem turned upside down.
- ›What the $999 actually buys is not stated anywhere. No number of agents, conversations, seats or volume. So the one published figure has no unit attached to it.
- ›Two higher tiers are named in that same description, covering agents built into your own product and trials the company sets up for you. Neither is priced, and both imply the company doing work rather than you signing up.
- ›There is a 14 day free trial, and nothing that loaded says what it includes or whether a card is needed.
How the price works
What you are charged for, and what makes the bill go up.
One published figure and no metering basis. The pricing page bails out to client side rendering and serves no plan cards, no tier names and no figures in its readable body. The price exists in the served document in two machine readable places only: the page description tag, which names the entry offering at $999 monthly and names two further tiers, and a structured data offer object which repeats $999 against the product.
The two upper tiers are named in the metadata without figures, described as covering in application agents and provisioned trials.
No unit accompanies the $999. Nothing published states whether it covers one agent or several, a volume of conversations, a number of seats, a number of resolved interactions or an account. For an autonomous conversational agent, where the category norm is metering by conversation or by outcome, a flat monthly figure without a stated allowance cannot be modeled against expected usage.
A fourteen day free trial is offered and is the only commercial term rendered to a reader. No trial scope, card requirement, seat minimum, contract length or annual option was located.
The same document publishes a machine callable agent gateway address, indicating the product is built for consumption by other automated systems.
What the contract says about your data
What the vendor commits to in writing once your data is in the product.
Not established. No security page, trust surface, processing agreement, sub processor listing, privacy policy or terms document was located from the pricing page or its navigation, and the page's client side rendering means any footer links may simply not have been present in the served markup.
That is a retrieval limitation rather than a finding, and it compounds the pricing limitation rather than standing alone: the same rendering behavior that hid the plan cards would hide a legal footer.
The custody question here is unusual and worth naming, because this vendor exposes an agent interface for other automated systems to call. That means the platform is not only holding conversation data from visitors to the customer's own website, it is providing a route by which external systems interact with that agent. A buyer should establish who may call that interface, how it is authenticated, what conversation content is retained from those calls, and whether transcripts of agent conversations with the customer's prospects are used to improve models beyond that customer's account.
None of that is addressed by anything reached here, and for a product whose entire function is autonomous conversation with a buyer's prospects, transcript retention and model training terms are the first questions rather than the last.
Getting started
What it costs and what is included before the product is running.
Not published. No setup fee, onboarding charge, migration rate, professional services rate, seat minimum or contract length was located, because the page body that would carry them does not render.
A fourteen day free trial is offered and is the only commercial term visible to a reader on the page. Its scope, limits and whether a card is required are all unstated in anything served.
One term is recoverable from the metadata alongside the price and is worth recording because it implies a services component: the upper tiers are described as covering in application agents and provisioned trials. A provisioned trial means the vendor stands something up on the customer's behalf rather than the customer self serving, which is a staffed step even where it is not separately charged. An in application agent means deployment inside the customer's own product rather than on a marketing site, which is an integration project rather than a script tag.
So the shape implied by the metadata is a self serve entry offering at a published rate and two higher tiers that involve the vendor doing work, with none of that work priced.
A buyer should establish what the entry figure includes in conversation volume before comparing it to anything, since a fixed monthly rate for an autonomous conversational agent is unusual in a category that generally meters by conversation or by resolution.
What to watch for
Where this pricing can surprise a buyer who has not read it closely.
The vendor your brief named as precedent, and it behaves exactly as described: the price is in the page description tag and nowhere a reader can see it.
The pricing page serves a heading, a trial call to action and a marker indicating the page has bailed out to client side rendering. No plan card, no tier name and no figure appears in the rendered body. What the served document does carry is the page description tag, and that tag states the entry offering at $999 monthly by name, together with the names of the two tiers above it. A structured data offer object elsewhere in the same document repeats the $999 figure against the product.
So the only two places this vendor's price exists in machine readable form are the metadata and the structured data, which are precisely the surfaces a search engine and an answer engine read, and precisely the surfaces a human never sees. That is the inverse of the ordinary staleness problem in this category, where a metadata figure lags the body. Here there is no body figure to lag.
It makes the vendor unusually well positioned to be quoted accurately and unusually badly positioned to be understood. An answer engine asked what this costs will return $999 monthly with confidence. A buyer landing on the page will see a trial button and no price at all.
The structural note that makes this vendor interesting beyond its pricing is that it publishes an agent gateway address in the same document, a machine callable route for other automated systems to converse with its agent. So this is a vendor building deliberately for machine consumption on the product side while its price reaches machines only by accident of metadata. Those two facts sitting in one document is the sharpest illustration in this index of how unevenly vendors have thought about being read by software.
What cannot be established from anything served:
- ›what the $999 buys in terms of agents, conversations, seats or volume
- ›what the two named tiers above it cost
- ›whether any of it is per month per agent or per account
- ›and what the fourteen day trial includes.
The metering basis is entirely absent, so the one published figure floats without a unit.
The numeric field carries $999, taken from the vendor's own metadata and corroborated by its own structured data.