SalesCaptain

The agents are the product's reason for existing and the pricing proves it. A phone agent answers inbound calls with a synthetic voice, qualifies the caller, answers questions and books an appointment. A chat agent does the same across text, webchat and social channels. Both are built in a no code builder and grounded on a knowledge base the operator supplies covering hours, policies and common questions, with templates provided per use case.

The economics confirm the placement rather than contradict it. Model usage is metered explicitly: phone agent minutes at twelve cents on the free tier and ten cents within the paid allowance, text generations at one cent each, and the paid tier's headline inclusion is two hundred agent call minutes. Pricing the model by the minute means the vendor's revenue moves with model usage.

What holds this off the top band is the substantial deterministic platform underneath. A phone system with call flows, interactive voice response and routing, a unified inbox, texting and rich messaging, automations, reviews, payments and appointment apps all function with no agent enabled, and several of them are the reason a local business buys a communications platform at all. The agents sit on top of working telephony rather than replacing it.

Ask what the agent does when the knowledge base has no answer, and whether it can be scoped to specific call types.

Autonomy is high and genuinely bounded, and the bounding mechanism is the strongest governance feature on this record.

Agents act without a human present: they answer calls, hold conversations, qualify, filter spam and write appointments into a calendar. The constraint is that they are grounded on a knowledge base the operator builds, covering stated hours, product availability, return policies, scheduling rules and frequently asked questions. Grounding an agent on a controlled corpus rather than letting it improvise from a general model is the correct architecture for a business where a wrong answer about a policy becomes a commitment, and the vendor makes it the central setup step rather than an advanced option.

Deterministic control sits around it. Call flows and chat flows are configured by drag and drop, interactive voice response and custom routing govern where a call goes, and automation rules fire actions on triggers with the customer defining conditions. Handoff to a person is described with full conversation context attached, and everything lands in a shared inbox a human can take over.

What is undescribed is the failure boundary. Two passes located nothing stating what an agent refuses to answer, whether it can quote a price or make a commitment, whether it escalates on uncertainty rather than guessing, or what a supervisor can review after the fact.

Ask what the agent does off script, and whether transcripts are reviewable.

Two passes located no model, no provider, no version and no architecture for any component: not the voice agent, not the speech recognition, not the synthesis, not the chat agent, not the text generation.

The metering makes the omission commercially awkward as well as technically opaque. Text generations are billed at a cent each and agent calls by the minute, so a customer's bill is a direct function of model consumption, and the vendor never defines the unit. What counts as one generation, whether a long reply costs the same as a short one, and whether a retry is chargeable are all unstated, which means the customer cannot forecast the variable half of their bill.

The substantive gap concerns voice. This agent speaks to consumers on the telephone, and the vendor describes the voice as natural sounding, which is the whole point of the feature. Nothing states which synthesis provider produces it, whether call audio leaves the platform to reach a third party model, whether recordings or transcripts train anything, or what accuracy the recognition achieves on a noisy line, which is the ordinary condition of a call to a roofing contractor.

No language coverage statement was located either, which matters for a product sold into United States local businesses with multilingual customer bases.

Ask which provider processes call audio, and whether recordings or transcripts train any model.

Two passes across the home page, pricing, product pages and documentation located no quantified outcome of any kind. No customer count, no answer rate, no booking rate, no lead recovery figure, no revenue result, no named reference customer and no case study with numbers.

What exists is qualitative and second hand. Reviews on third party platforms are genuinely positive and specific about the workflow, describing a medical spa capturing after hours enquiries for injectables and laser services and a roofing contractor consolidating communication, and one reviewer confirmed current use by uploading billing evidence. Those are real users describing real behaviour, which is worth more than an invented testimonial, and they are still not outcomes.

The absence is pointed for this particular product because the value proposition is arithmetic. A platform sold on never missing a lead, with agents answering calls around the clock, necessarily computes how many calls the agent answered that would otherwise have rung out, how many became appointments and what that is worth. Every one of those numbers exists inside the system in order to render the dashboards the vendor markets, and none is published in aggregate.

Investor names are published, which speaks to funding rather than performance.

Ask for the median after hours call capture rate and appointment conversion across customers in your vertical.

Squarely applicable and substantially unaddressed, which is the combination this axis exists to catch.

The product does not merely receive. It sells text marketing as a priced line with five thousand message credits included at the paid tier and a published per credit rate beyond that, alongside rich messaging, marketing flows and phone agents that place calls. That is bulk commercial messaging to consumers in the United States, which is the most heavily regulated outbound channel there is.

Two dedicated passes across the home page, pricing, product pages and documentation located no compliance surface at all. Nothing on consent capture or proof, nothing on opt out keyword handling, nothing on quiet hours or calling time restrictions, nothing on do not call screening, and nothing on the carrier registration required before a business may send application to person messaging on a standard ten digit number. Registration is not optional and carriers levy per violation charges for evading it, so a local business signing up here is being sold a messaging capability without being told about the gate in front of it.

The consumer facing nature sharpens it further. A medical spa texting patients and a contractor texting homeowners are exactly the senders that draw claims, and the statutory damages are per message.

Ask who completes carrier registration, how opt outs are captured and enforced across channels, and what quiet hour controls exist.

The documentary set is present and more complete than many at this size. A privacy policy, terms and conditions and a refund policy are all published and linked from every page, and a documentation site is maintained. Publishing a refund policy at all is unusual and is credited.

One defect is directly checkable and worth stating precisely because it is the kind of thing nobody notices. The footer carries a link labelled for the California consumer privacy statute, and it resolves to the site home page rather than to any notice. For a vendor selling into United States local businesses whose end customers are consumers, a broken route to the state privacy notice is the specific failure that matters, because it is the link a consumer or their lawyer would follow.

What could not be located across two passes is the processor layer. No data processing agreement, no stated retention period for call recordings, transcripts or message history, no subprocessor list and no processing location.

The holdings make those omissions consequential. This platform accumulates recorded telephone conversations with consumers, transcripts of them, message threads, payment records through a built in payments app, and appointment histories, all belonging to third parties who contacted a local business and have no relationship with this vendor.

Ask for the retention period on recordings and transcripts, and for a working route to the state privacy notice.

Not applicable in the provider sense and rated accordingly rather than penalised. Two passes located no contact database, no enrichment product, no purchased records, no third party data sourcing and no lead marketplace. The vendor supplies software, not data.

Every record in a deployment originates with the customer's own business. Contacts arrive because a consumer telephoned, texted, opened a webchat or messaged a social account belonging to the local business, or because the business imported its existing customer list. That is inbound by construction, which is a materially different and lower provenance exposure than any prospecting vendor in this index, and a buyer inherits no upstream licence terms, redistribution restriction or consent question from the vendor.

One provenance question runs the other way and is unaddressed. The marketing module lets a customer text its contact base, and the platform imposes no described check on where an imported list came from or whether those consumers consented to marketing. A contact captured because someone rang to ask about opening hours has not thereby agreed to receive promotional messages, and the product moves that record from one context to the other without comment.

Contact records also arrive through eighteen named vertical operational systems, so provenance may sit in a booking or practice management platform the vendor never sees.

Ask what consent state travels with a contact imported from an integration.

The classic extraction exposure is absent. Two passes located no scraping utility, no social network data harvesting, no contact sourcing and no browser extension of that kind. The eighteen named integrations are with operational systems the customer already licenses, connected through a published marketplace with a page per integration, which is the sanctioned pattern.

Where exposure genuinely sits is in the messaging channels, and the vendor does not describe the boundary. Social channel messaging runs through the major social platforms' business messaging programmes, which impose their own rules on automated replies, response windows and promotional content, and nothing published states which programme the integration operates under or what an agent is permitted to send inside it. An automated agent replying to social messages at volume is precisely the behaviour those programmes police.

The telephony and messaging side carries the same undescribed boundary. Numbers are ported or hosted, extra numbers are sold monthly, and text marketing runs at volume, all of which sit under carrier terms and a messaging registration the vendor never mentions.

The consequence falls on the customer, because it is their business account, their number and their carrier registration that get restricted.

Ask under which social messaging programme the agent replies, and who holds the carrier registration for outbound texting.

Two dedicated passes located no security page, no trust centre, no security contact, no responsible disclosure route, no published incident history, no status page and no statement of any technical control. Not encryption, not access control, not monitoring, not business continuity. The site's entire published governance surface is a privacy policy, terms and a refund policy.

The knowledge base grounding described on the autonomy axis is a genuine safety control and it is credited there rather than double counted, but it constrains what an agent says rather than protecting what the platform holds.

What the platform holds is the problem. Recorded telephone conversations between consumers and a local business, transcripts of those recordings, message threads across four channels, payment records through a built in payments application, and appointment histories. In several of the verticals the integration list names, those conversations carry material a reasonable person would consider private: a caller to a medical spa discussing a treatment, a caller to a legal practice describing a matter.

Whether any of that trains a model, whether tenants are isolated, how long recordings persist and who can access them are all unstated.

The payments application raises card data handling, which is unaddressed on every retrieved surface.

Ask whether recordings or transcripts train any model, what the retention period is, and who at the vendor can access a customer's recordings.

Squarely applicable, because this product's core function is a synthetic voice answering a telephone call from a member of the public.

A consumer who telephones a local business and reaches an agent is having a conversation with software that the vendor markets as natural sounding. That person did not opt into an automated interaction, has no relationship with this vendor, and in many cases is calling about something consequential, an appointment, a quote, a legal enquiry. Two passes located no disclosure position anywhere: no statement that the agent identifies itself, no configurable disclosure line, no guidance to customers on whether their jurisdiction requires one, and no default script published.

Recording compounds it. Transcripts and call summaries are marketed capabilities, which means calls are captured, and a substantial number of United States states require the consent of both parties to record a telephone conversation. Nothing published describes a recording notification, whether one plays by default, or whether the customer must configure it. The obligation falls on the local business, and the vendor sells them the capability without flagging the duty.

Several states have also introduced disclosure requirements specific to automated callers, and the vendor takes no position on them.

This is not a contradiction like some records in this index, it is a silence at the exact point where the product touches the public.

Ask whether the agent identifies itself by default, and whether a recording notice plays before capture.

Genuinely strong and, unusually, the depth is in the choice of systems rather than the count. Eighteen integrations are published individually on a dedicated marketplace subdomain with a page each, and a third party listing puts the total above fifty. What distinguishes them is that they are vertical operational systems rather than generic productivity tools: legal practice management, field service dispatch, roofing estimation, home improvement, pest control, spa and salon booking, tour reservations, hospitality property management, and four separate retail point of sale, commerce and shipping platforms.

Integrating with a roofing estimation system is a deliberate act requiring domain understanding, and doing it eighteen times across unrelated verticals represents real investment. For the segment this product serves it is the correct integration strategy, because a local business runs its whole operation inside one of those systems.

A marketplace architecture extends it further with first party applications for reviews, payments, appointments, social planning, a leaderboard and webhooks, and a documentation site is published. Mobile applications ship on both stores.

The ceiling is the developer surface. A webhooks application exists and a general automation connector is listed, and two passes located no public interface documentation, no rate limits, no self serve credentials and no description of what writes back to a connected system.

Ask what synchronises back to your operational system and in which direction.

Two passes located no hosting provider, no region, no country, no residency option, no single tenant deployment, no customer managed keys, no subprocessor list and no transfer mechanism.

The corporate structure makes the question concrete rather than academic. The site footer states that the product is a division of a private limited company, which is an Indian corporate form, and the named investors are venture funds operating from India. The pricing page is headed for the United States market and sells United States telephone numbers, so the operating picture is an Indian company serving American local businesses.

That arrangement is entirely legitimate and it makes the unanswered residency question sharper, because the data crossing that boundary is recorded telephone conversations between American consumers and their local providers, along with transcripts, message content and payment records.

A medical spa recording patient enquiries and a legal practice recording matter descriptions both operate under sectoral confidentiality expectations, and neither can answer where those recordings rest or which jurisdiction's authorities could reach them.

The absence of any published subprocessor list compounds it, since telephony and speech processing are almost certainly performed by third parties whose identities are not disclosed.

Ask in which country call recordings and transcripts are stored and processed, and for the subprocessor list covering telephony and speech.

Two dedicated passes located nothing. No security page, no trust centre, no attestation, no certification, no auditor, no assessment period, no penetration test summary, no completed questionnaire, no subprocessor list, no status page and no security contact. There is no credential to apply the test to, because none is claimed anywhere, which is a different failure from an unnamed badge row and lands in the same place.

The navigation confirms it rather than merely omitting it. The footer offers products, applications, integrations, documentation, terms, a refund policy, a privacy policy and a link labelled for a state privacy statute that resolves to the home page. There is no compliance section at all.

Two product lines make this materially worse than a generic absence. The platform records telephone conversations between consumers and businesses, including in verticals where the content is sensitive by nature. And it ships a first party payments application, which places card data in scope for the payment card standard, and no attestation or scope statement for that appears anywhere.

The published documentation site shows the vendor is willing to write for customers, so this reads as a publication gap rather than proof of weak controls, and the practical consequence for a buyer is identical: nothing can be verified before contact, and a business subject to any sectoral review has nothing to submit.

Ask what attestations exist, and specifically what covers the payments application.

The granularity is excellent and internal contradictions keep it out of the top band.

What is published is unusually complete. A permanently free tier with unlimited users, then two paid tiers, and a usage table pricing every consumable to the cent: calling per minute, inbox texting and marketing per message credit, text generations per generation, agent calls per minute, extra numbers monthly at two different models, extra locations, search monitoring, and desk handsets as a one time charge. Support response times are published per tier at forty eight hours, two hours, and a named relationship manager, and publishing a support commitment as a tier differentiator is rare. A refund policy exists. A referral schedule is published including a percentage of an enterprise referral's bill, which most vendors keep private.

The contradictions are on the pricing page itself and a buyer will hit them. The paid tier renders at one hundred and fifty nine dollars in some blocks and one hundred and ninety nine in others on the same page. The top tier is simultaneously described as custom pricing and as three hundred dollars monthly onwards. And the paid tier is headlined as unlimited everything while the table two rows below caps marketing credits at five thousand and agent minutes at two hundred.

Unlimited claimed above a table that contradicts it is the kind of thing that erodes trust in the rest of the sheet, which is a shame given how good the rest is.

Ask which paid tier price is current, and what unlimited excludes.

Two passes located no export route, no file format, no termination assistance, no notice period and no statement of what happens to data when an account closes.

For a communications platform the decisive exit question is the telephone number, and the asymmetry here is stark. Porting or hosting a number into the platform is published as free at every tier, prominently, as the first line of the pricing table. Nothing anywhere describes porting a number out. A local business's telephone number is printed on its vehicles, its signage and every listing it holds, and losing it is not a migration inconvenience but a loss of the business's identity. A vendor that advertises free inbound porting and is silent on outbound porting has documented only the direction that suits it.

The accumulated data compounds it: call recordings, transcripts, conversation history across four channels, contact records, appointment history, review history and payment records. Nothing states whether any of it can be retrieved, in what format, or how long it is retained after cancellation.

A refund policy exists, which addresses money rather than data, and a webhooks application gives a technically capable customer a partial route to stream data out while the account is live.

Ask in writing whether your number can be ported out on termination and on what timeline, before porting it in.

The product sends at volume across two of the most tightly policed channels and publishes no discipline for either.

Text marketing is a priced line with five thousand credits included at the paid tier and a per credit rate beyond it, and rich messaging is a named product. Two passes located no carrier registration guidance for application to person messaging, no discussion of the trust score that determines a sender's throughput, no opt out keyword handling, no throttling, no delivery rate reporting and no bounced or blocked message handling. Registration is a prerequisite rather than a best practice, and unregistered traffic is filtered or surcharged by carriers, so a customer can buy credits here and find their messages simply not arriving.

On the voice side, the agent places and receives calls from numbers ported or newly issued, and nothing addresses caller identity authentication, spam labelling, number reputation or remediation when a number is flagged. A local business whose main line gets labelled as spam loses inbound calls, which is the exact outcome this product is sold to prevent.

One adjacent capability points the right way and is credited: the agent filters spam on inbound calls, which protects the customer's attention rather than their reputation.

Ask who completes messaging registration, what delivery rate is achieved, and what happens if your number is flagged.

The segment is unusually well evidenced, and the evidence is the integration list rather than a positioning statement. Eighteen named integrations cover legal practice management, field service dispatch, roofing estimation, home improvement, pest control, spa and salon booking, tour reservations, hospitality property management, and four retail commerce, point of sale and shipping platforms. Every one is the operational system of a local service or retail business. No integration on the list serves a business to business revenue team.

The pricing structure says the same thing. A permanently free tier with unlimited users fits a single shop where everyone answers the phone. A separately priced extra location line at the paid tier reveals multi site local operators as the growth path. Desk handsets are sold as hardware, which is a physical premises assumption. Reviews come from medical spa and roofing operators, and search monitoring is priced as a feature, which is a local visibility concern rather than a pipeline one.

This record therefore belongs in this index as go to market technology for local businesses, and a revenue leader reading it should understand that it is not built for their motion. Stating that plainly is more useful than filing it as a generic communications platform.

Geographic coverage is the gap. The pricing page is headed for the United States and shows no other market, no other currency and no supported country list, while the operating company is Indian. No language coverage statement was located for the agents.

Ask which countries are supported for numbers, and which languages the agents handle.