AI in sales examples: From prospecting to pipeline

AI in sales examples are most useful when they show what the AI actually does, what the rep still owns, and how the workflow gets built. According to HubSpot’s 2024 State of AI in Business and Sales, 64% of sales professionals save one to five hours per week using AI to automate manual tasks — but the value depends on how the workflow is structured, not on the model alone.

The workflows below require sales reps to make final approval decisions on AI-generated outputs before they affect customer interactions. Every example keeps the rep in the loop for approval and is honest about privacy, consent, permissions, and governance. Where HubSpot products are referenced, prebuilt agents are distinguished from custom agent builder workflows, and beta capabilities are labeled as such.

This guide covers six examples — prospecting, qualification, outreach personalization, meeting prep and call summaries, post-meeting follow-up, and pipeline review — each described through the same lens: input, AI action, rep action, approval point, setup, and outcome.

Table of Contents

What is AI in sales?

AI in sales — sometimes referred to as artificial intelligence in sales in more formal contexts — is the use of machine learning, natural language processing, and agentic systems to automate or assist tasks across the sales funnel.

AI in sales reduces repetitive work across the sales funnel—research, drafting, note-taking, CRM logging, and prioritization—enabling reps to spend more time on conversations and less on administrative overhead. HubSpot offers three AI capabilities for sales: Breeze Assistant answers questions within the workflow, prebuilt agents (like prospecting agent and deal agent) handle specific sales tasks, and custom agents can be built in agent builder to fit the team’s process.

The practical question of how to use AI in sales for a growing team usually comes down to two decisions: which workflows to automate first, and how to keep the rep in the loop on every output that touches a customer.

Examples matter more than category definitions for sales teams making adoption decisions. Broad overviews of AI in sales explain the landscape; this article works the other direction, showing what six specific workflows look like when built and run. Each example is described using the same six-part lens — input/trigger, AI action, rep action, approval point, setup, and outcome — so a reader can lift the pattern directly into their own team.

What AI Can and Cannot Do in a Sales Workflow

AI is well-suited to bounded, repeatable tasks with clean inputs:

• Research summaries
• First drafts of outreach
• Meeting transcription
• Deal-level pattern surfacing
• Structured CRM updates from unstructured conversation

In these areas, the AI produces a first pass that a rep reviews, adjusts, and approves.

AI is not well-suited to judgment calls that depend on relationships, context outside the CRM, or ethical trade-offs. Deciding whether to walk away from a deal, when to escalate to a customer’s leadership, how to handle a sensitive negotiation moment, or whether a prospect’s needs actually fit the product remain human calls. Every workflow in this article keeps those decisions with the rep and reserves AI for the surrounding preparation, drafting, and documentation.

Consent, privacy, and permissions apply throughout. Recording a call requires the participants’ consent under the applicable jurisdiction. AI features that process customer data must comply with the data-handling agreements the customer signed. Access to AI capabilities depends on the account’s seat, tier, plan, and credit configuration. These constraints are noted alongside each product reference in the sections that follow.

AI in Sales Examples for Prospecting and Research

In prospecting, AI shortens the time between identifying a target account and sending a relevant first touch. AI reads public signals, enriches the CRM record with structured context, and drafts an outreach message the rep reviews before sending. The rep still owns targeting, message quality, and consent, but the AI compresses the manual research that once preceded each touch.

HubSpot’s prospecting agent helps teams research prospects and identify contacts. Teams that need a custom variation on the prospecting workflow (a different signal set, a different research source, or a different message pattern) can build one in agent builder rather than trying to force HubSpot’s prebuilt prospecting agent into a workflow it wasn’t designed for.

The evaluation question for sales reps is not “which is the best AI tool for sales prospecting”; rather, it’s which combination of prebuilt and custom agents fits the team’s actual process.

How AI Supports Sales Prospecting

AI supports sales prospecting by automating account research, sourcing contacts against the ideal customer profile (ICP), and drafting personalized outreach for rep review. The manager’s role is to define the ICP, approve the signal sources, monitor draft quality via sampling, and hold reps accountable for the send decision on every message. Answering how to use AI for sales prospecting well starts with the same discipline — the manager owns the guardrails, the rep owns the send.

Example workflow:

• Input/trigger: A new named account is added to the target list, or a defined business signal (funding round, leadership change, product launch) fires on an in-ICP account.
• What the AI does: Sources verified contact details, summarizes recent public activity, matches use case to the account context, and drafts a first outreach message anchored to the signal.
• What the rep or manager reviews: The rep validates contact accuracy, edits the draft for voice and specificity, and confirms that the send is appropriate. Manager samples message quality weekly.
• Approval or escalation: The rep approves every send. Escalate to the manager when a draft references sensitive data or signals that the rep cannot verify.
• Setup: Enable prospecting agent from Agent Hub, define the ICP filters and signal sources, and confirm seat, credit, and permission requirements against current HubSpot documentation. For teams that need a custom variation, agent builder configures the same input/output pattern without code.
• Success metric (outcome): Reply rate on AI-drafted-and-reviewed sends versus reply rate on manual sends over the same 30-day window.

AI in Sales Examples for Lead Qualification and Routing

AI in lead qualification combines firmographic fit, in-CRM engagement history, and product usage signals to prioritize new leads and route them to the right rep faster. AI produces a recommendation; the rep confirms fit and owns the next step. Lead qualification workflows use fit, engagement, and CRM context without replacing rep judgment — a scored lead is a prioritization signal, not a verdict.

HubSpot Smart CRM provides connected customer and deal context for sales workflows, which is what makes qualification useful in the first place. Layered on top of it, AI Guided Selling surfaces next-best actions on existing opportunities, and Sales Automation handles the routing, task creation, and follow-up sequencing that turns a qualified lead into a conversation. Reps can also read the more tactical qualification pointers in HubSpot’s inbound sales guide for the interpersonal layer that AI cannot substitute for.

How AI Qualifies and Routes Leads

AI qualifies and routes leads by scoring fit against the ICP, weighting engagement history, and matching the lead to the appropriate rep or queue. A scored lead is not a qualified lead — it is a prioritization signal that a rep uses to sequence outreach. Routing based solely on model output, without any rep confirmation, is not recommended.

Example workflow:

• Input/trigger: A new lead enters the CRM through a form fill, event signup, product signup, or third-party data source.
• What the AI does: Scores the lead on fit and engagement, matches against the routing rules, and creates an assigned task for the appropriate rep.
• What the rep or manager reviews: The rep confirms the disposition on first touch and adjusts the record if the model’s fit assessment appears incorrect. Manager reviews model drift weekly against actual conversion outcomes.
• Approval or escalation: The rep owns the disposition. Escalate to RevOps when scoring appears systematically off or when routing rules produce ownership conflicts.
• Setup: Configure lead-scoring criteria and routing rules in Sales Hub; align them with the ICP definition. A custom qualification agent can be built in Agent Builder to layer additional CRM-native scoring logic on top of the standard rules, subject to access documentation.
• Success metric (outcome): Time-to-first-touch on newly created leads and downstream conversion rate by score band.

AI in Sales Examples for Outreach Personalization

AI-drafted outreach personalization uses connected CRM context — company details, past interactions, engagement history, industry signal — to produce a personalized first draft that the rep reviews before sending. Personalization improves reply quality when the underlying signals are accurate, and the rep confirms the message reads like something a person would actually write.

Consent and appropriate-use expectations sit around this workflow. Every draft should reference only information the CRM is authorized to hold, respect the recipient’s known preferences and opt-outs, and pass a rep’s review before sending. Automated sends without rep approval are not recommended for cold outreach; they remove the human review step and increase both trust and compliance risks.

How AI Drafts Personalized Outreach

AI drafts personalized outreach by combining a template pattern with contact- and account-specific context pulled from the CRM. AI drafts should still be edited before sending. Reps who paste AI drafts directly into email without editing lose most of the personalization benefit because the draft still reads generically to a buyer who receives similar messages every week.

Example workflow:

• Input/trigger: A rep decides to reach out to a contact who has engaged with content, attended a webinar, or matches a signal-driven priority list.
• What the AI does: Pulls the relevant CRM context, retrieves any recent public signal, and drafts a personalized email that references the specific context.
• What the rep or manager reviews: The rep edits the draft to match the voice, verifies the accuracy of every claim, and confirms the reference actually lands. Manager reviews samples for tone drift.
• Approval or escalation: Rep approves every send. Escalate to manager or legal when the message touches sensitive customer data or regulated categories.
• Setup: Enable AI drafting inside Sales Hub email templates, connect the CRM properties the draft should read, and define voice/brand guardrails. A custom outreach personalization agent can be assembled in agent builder when the standard drafting flow needs additional inputs like product-usage or intent signals.
• Success metric (outcome): Reply rate on AI-drafted-and-edited sends compared to fully manual and fully templated sends over a controlled 30-day window.

AI in Sales Examples for Meeting Preparation and Call Summaries

AI meeting prep and call summarization compresses the time between scheduling a meeting and having a clean, actionable record of it. Before the meeting, AI assembles the account context, recent activity, and any relevant prior notes. During or after the meeting, AI transcribes the conversation, drafts a summary, and extracts proposed next steps for rep review.

HubSpot’s Notetaker joins Zoom, Google Meet, and Microsoft Teams meetings to record, transcribe, and summarize them, and it can also work from Zoom or Teams transcripts captured in the meeting record. When paired with Smart Deal Progression, HubSpot’s Notetaker adds meeting prep beforehand and, afterward, a summary, action items, and suggested CRM updates. Reps should review the summary, next steps, and suggested updates before accepting them into the CRM.

How AI Prepares Reps and Summarizes Calls

AI prepares reps by assembling context that would otherwise take 10-15 minutes of manual research; AI summarizes calls by turning the transcript into a structured recap with proposed next steps. In both cases, the rep is the final author — the AI produces a first draft, and the rep confirms it before the record is stored.

Example workflow:

• Input/trigger: Meeting is scheduled on the rep’s calendar with an associated CRM contact, company, or deal record.
• What the AI does: Before the meeting, it assembles account context, prior interactions, and open items. During/after the meeting, it transcribes the call (with consent), summarizes the discussion, and drafts proposed next steps.
• What the rep or manager reviews: The rep or manager validates the summary, edits the proposed next steps for accuracy, and approves it before the summary is saved to the CRM record. The manager can review deal-level summaries to spot missed follow-through.
• Approval or escalation: Rep approves the summary and any CRM property updates before they persist. Escalate to security when a transcript captures sensitive data that should not be stored.
• Setup: Enable HubSpot’s Notetaker or BYO transcripts via connected apps (Zoom, Microsoft Teams, Google Meet); confirm calendar and email sync; audit consent and transcription enablement for each rep. A custom pre-meeting briefing agent can be built in Agent Builder to layer segment-specific context onto Notetaker’s default prep.
• Success metric (outcome): Reduction in post-meeting admin time per rep, combined with an increase in downstream deal-record completeness (a proxy for cleaner handoffs).

AI in Sales Examples for Post-Meeting Follow-Up

AI for post-meeting follow-up turns the meeting outcome into a concrete next-step package: a follow-up email draft, proposed CRM updates, and a task list. This is the only follow-up-agent example in this article, and it is intentionally scoped: the AI produces the artifacts, the rep approves the sends and property changes, and exceptions escalate to the manager.

HubSpot’s deal agent turns customer interactions and pipeline signals into actionable recommendations, guidance, and ready-to-run workflows that move deals forward. Deal agent reads from HubSpot’s Notetaker and BYO transcripts, plus email sync and the CRM context layer. It then acts on the rep workflow in the CRM and inbox with next-best actions, email drafts, and CRM/methodology updates.

How AI Turns Meeting Outcomes Into Next Steps

AI turns meeting outcomes into next steps by combining the call transcript, associated email activity, and CRM state into a proposed next-step package for the rep to review. The rep approves each element — the follow-up email draft, the CRM property updates, the tasks — before any change is committed. Deals with unusual signals (a customer objection the model does not understand, a regulatory reference, a request for legal review) should route to the manager rather than being auto-processed.

Example workflow:

• Input/trigger: A meeting ends, and its transcript, associated emails, and deal-record state become available.
• What the AI does: Drafts a follow-up email tied to the discussion, proposes CRM property updates (stage, next step, close date), and creates a task list with owners and dates.
• What the rep or manager reviews: The rep reviews each draft and property change; edits before approving; adjusts task ownership if needed. Manager samples deal-level updates weekly.
• Approval or escalation: The rep approves drafts and CRM updates. Escalate to manager when the meeting surfaced a legal/security/pricing exception or when the AI’s summary conflicts with the rep’s own recall.
• Setup: Enable deal agent with transcript sync via Notetaker or connected apps, confirm Sales Hub / Service Hub Pro+ seat and credit availability, and define which CRM properties require rep approval versus auto-apply. For custom follow-up patterns unique to the team’s segment or motion, configure a custom agent in agent builder.
• Success metric (outcome): Time from meeting end to follow-up send, plus follow-through rate on proposed next steps in the seven days following the meeting.

How to Configure a Post-Meeting Workflow in agent builder

Teams that need a custom variation on the follow-up workflow can build one in agent builder. Agent builder enables teams to configure custom GTM agents without code. A configuration outline for a custom post-meeting follow-up agent looks like this:

• Define the trigger — a meeting ends on a deal that meets specified criteria (segment, stage, deal size).
• Define the inputs the agent can read — meeting transcript, associated emails, CRM record, sales content library.
• Define the outputs the agent proposes — email draft, CRM property updates, task list.
• Define the approval steps — which outputs require rep review, which allow auto-apply, which escalate to manager.
• Define exceptions — deal characteristics that route the workflow to a human owner rather than continuing agent processing.
• Define the metrics that will judge whether the custom agent is working before rolling it out beyond a pilot group.

Agent builder and the broader agent ecosystem live under Agent Hub — the central home for activating and managing AI agents across the sales org, including deal agent and prospecting agent.

AI in Sales Examples for Pipeline Review and Forecasting

AI supports pipeline review by summarizing deal-level state, surfacing stall-risk signals, and pointing managers at the deals that most need intervention. AI does not guarantee forecast accuracy in pipeline review. Forecast accuracy remains a function of the underlying CRM data quality, the sales methodology, and the rep and manager judgment applied at the review point — AI can make the review faster and more consistent, not more certain.

Teams can use AI guided selling for suggested tasks and next-best actions based on buying signals, and deal pipeline software to track deals across stages in real time, with predictive deal scoring to surface risks and opportunities. Sales Hub also supports sequences, tasks, and workflows, and pipeline performance can be reviewed in dashboards and reports.

Reps who want the broader context can also read the sales process guide, which grounds pipeline review inside the stages the AI is analyzing.

How AI Surfaces Pipeline Risks and Summaries

AI surfaces pipeline risk by looking for patterns across engagement, activity, and CRM signals such as:

• Deals that have stalled beyond typical cycle length
• Deals that lost the champion contact
• Deals that lack recent multi-threaded engagement

The manager reviews the surfaced list, decides which deals warrant intervention, and makes the final forecast decision. The AI’s job is to compress the manager’s review time, not to make the forecast call.

Example workflow:

• Input/trigger: Weekly pipeline review meeting is scheduled, or a manager opens the pipeline dashboard between reviews.
• What the AI does: Summarizes deal-level state, flags stall-risk deals with the signals that produced the flag, suggests candidate interventions.
• What the rep or manager reviews: Manager confirms interventions, updates forecast category calls, and assigns follow-through to reps. Rep contests flags when the AI lacks context.
• Approval or escalation: Manager owns the forecast and intervention decisions. Escalate to CRO when the surfaced risk crosses a threshold (e.g., a specific dollar value at risk in the current quarter).
• Setup: Enable AI guided selling and confirm the deal-health signal configuration inside Sales Hub; ensure pipeline stages, activity capture, and CRM property definitions are current. A custom pipeline-summary agent can be built in agent builder to surface segment-specific risk signals the standard configuration does not cover.
• Success metric (outcome): Review meeting time per deal, plus the accuracy of the manager-owned forecast against actual close in the following period.

Your 30-Day Pilot Plan

A 30-day pilot is the fastest way to move from AI-in-sales awareness to informed adoption. Start with two HubSpot workflows that produce measurable outcomes within a month — Lead Qualification and Post-Meeting Follow-Up — before expanding to higher-touch workflows such as personalized outreach and pipeline review. The plan below is a starting checklist, not a finished project plan; adapt it to the team’s actual constraints such as CRM data quality, rep enablement capacity, and manager review bandwidth.

• Week 1 — Define owners and baselines. Name a single owner for the pilot. Capture 4-week baselines on the two chosen workflows (time-to-first-touch, follow-up send time, downstream conversion or follow-through rate).
• Week 1 — Enable and configure. Turn on the relevant prebuilt agents in a controlled group. Verify seat, credit, and permission requirements against current HubSpot documentation. Confirm consent and privacy configuration.
• Week 2 — Train the pilot group. Walk each rep through the approval points, escalation rules, and manager review cadence. Provide written guidance on what AI drafts to accept, edit, or reject.
• Week 2 — Run the workflows live. Log every AI output the rep edits or rejects so the pattern is visible in the retro.
• Week 3 — Weekly review. Manager samples AI-produced work, discusses edits, and updates the guidance document based on what the pilot is showing.
• Week 4 — Compare to baseline. Measure the two chosen workflows against the pre-pilot baseline. Decide stop/iterate/scale.
• Stop criteria — the workflow produces work reps must rewrite from scratch, or the AI output creates trust or compliance risk. Retire the workflow.
• Iterate criteria — the output is directionally useful but needs process changes (better inputs, tighter approval rules, new guardrails). Adjust and re-baseline for another 30 days.
• Scale criteria — the workflow beats baseline on the defined success metric, reps report time savings, and manager sampling shows consistent quality. Roll out to the next group and add a third workflow.

Frequently Asked Questions About AI in Sales

Do I need engineering resources to build these workflows?

No. Prebuilt agents like prospecting agent and Notetaker are set up in the HubSpot UI. Prospecting agent requires Super Admin or Access prospecting agent permission, and Notetaker needs a connected calendar and call recording and transcription enabled. With agent builder, users with Super Admin or agent builder permission can create custom agents by writing instructions in plain language, with no code. Engineering help may still be useful for integrations beyond HubSpot’s built-in connections and Model Context Protocol (MCP) support, or for using agent output outside HubSpot’s platform.

How long does it take to launch an AI sales workflow?

Timelines vary meaningfully by workflow complexity, the state of the underlying CRM data, and the level of configuration a team wants before rollout. The 30-day pilot plan above is a reasonable starting rhythm for two focused workflows in a controlled group; broader rollout across all reps typically takes longer and depends on data cleanup, enablement, and manager review capacity. Avoid claims of universal launch speeds — the honest answer is that clean data and clear approval rules matter more than model speed.

How do we keep AI output accurate and on-brand?

Accuracy and voice are functions of three things: the quality of the CRM data the AI reads, the specificity of the instructions and templates the AI is configured with, and the discipline of the rep review step. When a manager samples AI-produced work weekly by reading a random subset and giving structured feedback, this practice catches drift before it becomes systemic; updating templates and guardrails based on that feedback keeps output aligned to the team’s voice.

Will AI replace sales reps?

No. The AI in sales examples in this article are structured to keep the rep in the loop on every decision that shapes the customer relationship — targeting, message quality, qualification, negotiation, and the send/approve step. The tasks AI handles well (research, drafting, summarizing, structured CRM updates) are the ones that consumed rep time without producing customer conversations. Reallocating that time to conversations is the intended effect.

Can these workflows extend to marketing and service?

Yes. AI in sales and marketing overlaps meaningfully — the same prospecting, personalization, and pipeline-signal workflows have marketing equivalents in campaign personalization, engagement scoring, and attribution. Service teams use similar patterns for triage, response drafting, and case summarization. HubSpot’s coverage of sales innovation covers the broader theme, and specific service and marketing use cases are outside the scope of this article.

Making AI in Sales Examples Repeatable

The six workflows above are useful examples because each is specific about what the AI does, what the rep still owns, and how the workflow gets built and reviewed. Successfully adopting AI in sales is less about picking the most impressive model and more about defining approval points, sampling AI-produced work, keeping consent and permissions explicit, and running a 30-day pilot before scaling. The skill the article implicitly builds is the ability to evaluate any new AI capability against the same six-part lens.

For teams looking to sharpen the underlying capabilities that make AI adoption stick, HubSpot’s guides to personal selling and how to improve sales skills cover the human layer that determines whether AI actually produces better outcomes.

In my own advisory work with sales and RevOps leaders, I have consistently seen the teams that succeed with AI in sales start narrow — one or two workflows, one pilot group, one manager reviewing edits weekly — before scaling to the rest of the org. The teams that struggle usually enable too many AI features at once, skip the approval points, and end up with output no one trusts. Narrow, disciplined, and measured beats broad and enthusiastic every time.