AI Sales Representatives for Inbound Lead Conversion

showme · Tech · 2025 · source Agentic: yes · promotional bias: low · source-backed · extracted with prompt 0.5.4 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: Inbound sales conversion

ShowMe wanted to convert inbound website visitors with human-like sales engagement, but human sales coverage was too expensive for low-intent or unqualified leads.

Constraints:

inbound visitors converted better when a human sales rep explained the product
many visitors were too early or unqualified to justify human sales time
voice and video conversations could not tolerate high latency
sales agents needed customer-specific sales guidance, objections, and product demos
customers needed visibility into conversation quality before trusting the agent

Architecture

ShowMe built AI sales representatives as digital teammates with sales skills and operational skills. Conversation agents handle website/video/voice/text interactions and product demos; workflow orchestration decides follow-up actions such as emails, calls, handoffs, and CRM updates; evaluator agents score call outcomes and create structured fields; creator agents turn customer documentation, sales transcripts, pitch decks, and training material into prompt guidance for customer-specific agents. Slack and CRM reporting make the agent look and operate like a teammate rather than a separate admin tool.

Diagram confidence: reconstructed (reconstructed from prose).

flowchart LR
  inbound-lead-channel["inbound-lead-channel"] --> conversation-agent-sequence["conversation-agent-sequence"]
  sales-guidance-knowledge["sales-guidance-knowledge"] --> conversation-agent-sequence["conversation-agent-sequence"]
  demo-tooling["demo-tooling"] --> conversation-agent-sequence["conversation-agent-sequence"]
  conversation-agent-sequence["conversation-agent-sequence"] --> workflow-orchestration-layer["workflow-orchestration-layer"]
  conversation-agent-sequence["conversation-agent-sequence"] --> evaluator-agents["evaluator-agents"]
  workflow-orchestration-layer["workflow-orchestration-layer"] --> orchestrator-agent["orchestrator-agent"]
  workflow-orchestration-layer["workflow-orchestration-layer"] --> crm-and-communication-systems["crm-and-communication-systems"]
  workflow-orchestration-layer["workflow-orchestration-layer"] --> human-sales-team["human-sales-team"]
  customer-source-materials["customer-source-materials"] --> creator-agents["creator-agents"]
  creator-agents["creator-agents"] --> sales-guidance-knowledge["sales-guidance-knowledge"]
  evaluator-agents["evaluator-agents"] --> customer-quality-review["customer-quality-review"]
  crm-and-communication-systems["crm-and-communication-systems"] --> customer-quality-review["customer-quality-review"]

Drag to pan · scroll or use +/− to zoom · reconstructed from the source's described component topology

Components

Component	Role	Interacts with
`inbound-lead-channel`	Website, phone, video, WhatsApp, email, or text entry point for prospects.	`conversation-agent-sequence`
`conversation-agent-sequence`	Multiple voice/video/text agents for greeting, discovery, qualification, pitching, product exploration, and handoff decisions.	`inbound-lead-channel`, `sales-guidance-knowledge`, `demo-tooling`, `workflow-orchestration-layer`, `evaluator-agents`
`sales-guidance-knowledge`	Customer-specific sales skills, value propositions, ICP rules, objections, and process guidance built from calls and enablement documents.	`conversation-agent-sequence`, `creator-agents`
`demo-tooling`	Product slides, screenshare, video segments, and product-demo actions shown during sales conversations.	`conversation-agent-sequence`
`workflow-orchestration-layer`	Stateful workflow layer that chooses next sales actions after each conversation outcome.	`conversation-agent-sequence`, `orchestrator-agent`, `crm-and-communication-systems`, `human-sales-team`
`orchestrator-agent`	Planned flexible agent that can break out of predefined follow-up paths when a lead requests a different timing or action.	`workflow-orchestration-layer`, `crm-and-communication-systems`
`evaluator-agents`	Extract call outcomes, custom fields, sentiment, handoff needs, and quality signals from conversations.	`conversation-agent-sequence`, `workflow-orchestration-layer`, `customer-quality-review`
`creator-agents`	Generate prompts and guidance artifacts for customer-specific conversation agents from documentation and transcripts.	`sales-guidance-knowledge`, `customer-source-materials`
`customer-source-materials`	Sales transcripts, pitch decks, enablement documents, ICP spreadsheets, product docs, and training materials provided by customers.	`creator-agents`
`crm-and-communication-systems`	Slack, email, CRM, HubSpot, Salesforce, and dashboards used for reports, follow-ups, and visibility.	`workflow-orchestration-layer`, `orchestrator-agent`, `customer-quality-review`
`customer-quality-review`	Customer review loop for conversation quality, feedback, tests, and rollout confidence.	`evaluator-agents`, `crm-and-communication-systems`
`human-sales-team`	Receives handoffs for complex or high-value leads and provides training examples for agent behavior.	`workflow-orchestration-layer`, `customer-source-materials`

Data flow

A lead enters from an inbound channel such as a website interaction. A conversation agent speaks, texts, shows slides, demos product areas, qualifies the prospect, and records the outcome. Workflow logic decides deterministic next actions such as an email or next-day call, while the planned orchestrator agent adds flexibility for off-happy-path timing and personalization. Evaluator agents extract call fields, sentiment, handoff needs, and next-step signals; creator agents ingest sales-team transcripts and enablement material to build customer-specific sales guidance. Reports and CRM/Slack updates expose quality to the customer.

4 supporting quotes

“they can share their screen. They can speak via voice via text. They can also demo a product.”transcript, stated

“These workflows is like a layer that is above the these agents and that basically have both deterministic and indeterministic actions.”transcript, stated

“we have agents evaluating the outcome of the call and understanding what's the next path that in a deterministic way it has to pick inside the workflow.”transcript, stated

“we're getting the reports of our agents and we're and keeping it tidy and good looking for our customers and we send them.”transcript, stated

Why this architecture

ShowMe decomposed the sales representative into several agents and a workflow layer because live voice sales cannot tolerate heavy-model latency and because one large agent with every tool became confused. Deterministic workflow paths prevent hallucinated follow-up behavior, while evaluators and customer reporting build trust before customers hand over all leads. The teammate metaphor also shaped the UX: customers interact through Slack, reports, and CRM rather than a separate back-office prompting tool.

4 supporting quotes

“Why we did this is basically a limitation of the current models and a trade-off of latency.”transcript, stated

“if we had all the tools for one single agent h it it got into a mess and sometimes it was calling the tool that it was not the correct one.”transcript, stated

“because we don't want like for the moment any hallucination in that part.”transcript, stated

“What we want to this to look like is that you have a co-orker and you communicate with it as you communicate with other co-workers.”transcript, stated

Models used: voice agents · video agents · RAG knowledge base · evaluator agents · creator agents · orchestrator agent

Scale

ShowMe’s first MVP was built in about two weeks.
“we went over did the product in two weeks that was basically you could you had some videos h you could select which video you wanted to see from the product and then there was an agent that was going a voice agent”transcript, implied
The workflow handled about 70% of cases end-to-end and handed off around 30% to humans at the time described.
“70 70% of the times we're able to handle the workflow till the end 30% we have to hand off”transcript, stated
Customer rollout typically took one to two months before full lead coverage.
“usually it's a thing of one month two months until they are 100% rolled out on all of their leads.”transcript, stated
The post-onboarding evaluation process can reduce review to almost 5% of conversations.
“then it continues reducing until it's almost like 5% of the conversation.”transcript, stated

Alternatives considered (roads not taken)

Use human sales reps for every inbound visitor

Rejected because: Human engagement converted but was too expensive for visitors who were early or unqualified.

“we're putting humans into this but it's costing us a lot of money that that we're investing in people that are not right”transcript, stated

Use free trials to solve inbound conversion

Rejected because: The previous product was complex and the aha moment came too late in the journey.

“free trial didn't really work for us and then on the other hand we started seeing how AI was getting better”transcript, stated

“we realized we had a pretty complex product and a product that the aa moment like was too late in the journey”transcript, stated

Use one large conversation agent with every tool

Rejected because: The all-in-one agent became messy, forgot earlier questions, and sometimes called the wrong tool.

“the agent was forgetting what they started asking at the beginning. So it started to pitch and then again started to ask the same questions”transcript, stated

Train from only the best account executive’s calls

Rejected because: Using only top-AE calls missed the diversity of how the sales team actually sells.

“give us the calls the five best calls from your best AE so we were basically selling like their best AE”transcript, stated

“then we were not getting the rest of them right so this is one of those realizations that that was key here”transcript, stated

Patterns

`multi-agent-orchestration` (agent-architecture)

ShowMe divides sales work across conversation, evaluator, creator, and planned orchestrator agents rather than one universal sales agent.

“we have the conversational agents, evaluators, and creators. This is more more or less how we call them.”transcript, stated

`workflow-decomposition` (workflow)

The sales funnel is represented as workflow steps and smaller agent responsibilities such as greeting, qualification, pitching, and follow-up.

“you need to divide in different agents actually inside it with different prompts and one is going to take over the the greetings and first discovery.”transcript, stated

`tool-use-orchestration` (agent-architecture)

Sales agents use demos, slides, calls, text, CRM actions, and workflow tools chosen according to the sales stage.

“here you can have in the end, hey, I have to pull up this slide. I have to pull up this other slide.”transcript, stated

`human-in-the-loop-gate` (safety)

The system hands off dead ends and high-uncertainty cases to human sales staff instead of forcing the agent through every path.

“when we're not able when we get to an end to that dead end what happens is that that we hand off to the human”transcript, stated

`eval-driven-development` (evaluation)

Customer feedback creates tests that the conversation agent must pass before the team trusts the updated prompt behavior.

“Automatically a test is created and the agent try to go through that conversation again”transcript, stated

`confidence-thresholding` (evaluation)

After the initial review period, evaluator confidence highlights only the conversations that customers still need to review.

“with that evaluator confidence level that is giving we're getting basically or spotting for the customer which are the conversations that they need to evaluate”transcript, stated

Tradeoffs

Start manual and automate progressively

Chose manual-heavy MVP and staged automation over fully automated onboarding and operations from day one.

Rationale: The team validated demand quickly, then automated pieces of the workflow as the product matured.

“we started this is fake it till you make it and we almost everything we did it manual and then we continue automating”transcript, stated

Use deterministic workflow paths before adding more agentic flexibility

Chose deterministic next actions for follow-up over fully free-form agent decisions for every sales action.

Rationale: Determinism reduced hallucinated follow-up behavior while the team added orchestrator flexibility for cases that genuinely required it.

“the paths that can can get are deterministic. H and and basically because we don't want like for the moment any hallucination in that part.”transcript, stated

Roll out gradually with customers

Chose low-risk pilots, low-quality leads, and progressive rollout over immediate full replacement of inbound sales handling.

Rationale: Customers needed to test agents in live conditions and build confidence before routing all leads through the system.

“we try to find the best environment as for them to test our agents and our technology in the way that is the less the least riskiest possible.”transcript, stated

“Sometimes it's passing some of their leads, the lowest quality leads through this part”transcript, stated

Gotchas (what bit them)

Live voice sales could not use the heaviest possible model for every step.

Root cause: Voice conversations have strict latency expectations, and heavier models introduce too much delay.

Resolution: Split the conversation into smaller agents and use smaller models for latency-sensitive conversational paths.

“The heavier the model you cannot do in a conversation that is via voice. You cannot have high latency.”transcript, stated

“So that made us use smaller models and for that we needed to divide the problem”transcript, stated

Customers did not automatically trust that good agent conversations were happening.

Root cause: The product initially focused on conversation quality itself, but customers lacked visibility into that quality.

Resolution: Add reporting, Slack conversation sharing, CRM logging, dashboards, and review loops so customers can inspect quality.

“we know they are great but our customers don't really have the visibility into that.”transcript, stated

“sharing the conversations via Slack logging everything into the CRM and so on having dashboards with full visibility”transcript, stated

A fixed follow-up workflow could miss user-specific timing requests.

Root cause: The workflow originally followed predefined actions even when a prospect asked for a different call-back time.

Resolution: Build an orchestrator layer that can create new actions and personalize next steps beyond the default path.

“they want us to call them in two hours. I'm going to create a new action that is call them in two hours.”transcript, stated

“orchestrator is the one that is break with that I would say lack of flexibility”transcript, stated

Best-performer call samples were too narrow for customer-specific sales guidance.

Root cause: Only copying the best AE’s calls missed other successful team behaviors and edge cases.

Resolution: Ask for conversations from different salespeople plus the same training materials used to onboard human sales staff.

“we need to to get into different conversations from different people too”transcript, stated

“give us the same materials you used to train your sales team.”transcript, stated

Evaluation

ShowMe evaluates at two levels: business metrics such as conversion, qualified meetings, and sales outcomes, plus conversation-level customer review. During the first month, customers review nearly all conversations and their feedback becomes tests; later, evaluator confidence identifies the smaller set of conversations that still need human review.

Metrics: conversion · qualified meetings created · sales closed · customer-reviewed conversation quality · tests created from customer feedback · review load reduced toward about 5% of conversations

3 supporting quotes

“the success metrics the size is as probably every other human sales man it's basically conversion”transcript, stated

“whether you're making those qualified meeting that you promised and whether you're doing the sales that you promised.”transcript, stated

“almost every conversation is reviewed with the customer reviews almost all of the conversations that are happening.”transcript, stated

Outcomes

ShowMe created AI sales reps that conduct video, voice, text, demo, and follow-up interactions
the workflow reportedly handled about 70% of cases end-to-end at the time described
agent onboarding and reporting were partially automated through Slack and customer-visible workflows
customer feedback created test batteries for conversation-agent improvement

The transcript describes an early-stage 2025 company with production customers, gradual one- to two-month customer rollouts, and a stated goal of reducing 30% human handoffs as the orchestrator becomes more flexible.