Can Natural Language Interfaces Accelerate Sales Operations?
How This Started
At Fairmart, I found managing a sales team selling to SMBs had specific challenges, specifically around data quality, compliance, and prioritization. This is a product concept note to think through how we can improve tooling and experiences for SMB focussed sales teams, specifically focussing on the needs of the sales manager by reducing reporting burdens on their reps.
SMB Sales Ops is a High Friction Process
Modern SMB sales managers leverage CRM dashboards and pipelines to manage their teams at scale, but data quality is a fundamental impediment to decision quality. The majority of sales tools are designed for enterprise sales flows, and not optimised for SMB.
SMB segments typically have lower ACVs, relying on high deal velocity to drive revenue; decision speed is essential to success. The problem is especially pronounced with large SMB focused sales teams where interactions happen on WhatsApp or in person.
Typically managers review opportunities on a weekly basis with the team to assess overall pipeline, set individual priorities for sales reps, and identify key opportunities that require direct management support. These systems rely on quality and timely data in order to serve their purpose. Individual sales reps generally resist extensive CRM updates as they find the work laborious and superfluous. Even though actions may have been taken, they frequently are not updated into the CRM or they lack sufficient detail for managers to assess deal probability.
Implementing Natural Language to Streamline Sales Ops
Natural language interfaces are now a reality. We propose to implement a sales agent that takes natural language inputs and converts them into CRM record updates on behalf of sales reps. We use existing communications layers, such as WhatsApp, Slack or Google Workspace Chat, where communication already takes place. Sales reps are prompted daily to provide an update on the meetings and tasks they had scheduled on their calendars, and prompted to review updates for accuracy prior to writing to the CRM. If sales reps are on the move, they can even drop a voice note.
We support sales reps further by helping them plan their weekly tasks. Our planner agent reviews open deals, time of last activities, and proposes a realistic task list for the week, broken down by day. Sales reps provide feedback on proposed plans that surfaces client realities, and improves pipeline accuracy. For example, in response to a proposed follow up, a rep may say that a decision maker stated that they don’t have a budget this quarter; this information would then recommend the rep to demote a deal from immediate opportunity to future pipeline.
Sales managers are empowered when they have better data in real time, and do not spend time interrogating deal pipelines. Instead, they have a clearer picture on opportunity health and where their attention is needed most to support individual sales reps.
Through this process, we institutionalise the rhythms of your team, make process compliance a natural conversation, and surface non-compliance in real time.
What We Don’t Do
We explicitly do not do any externally facing content such as customer email drafts, or call scripts. Existing GTM tools handle this already and requirements vary greatly across sales teams. Our focus is solely on improving internal sales operations.
Build on Top of Your Existing Sales Stack
Our approach is to integrate into the existing sales stack of your team. We do not build a new app surface, but rather deploy to where your team communication lives today: Slack, Google Workspace, Teams, or WhatsApp, in both text and voice. Multichannel updates are supported as well, so sales reps and managers can provide or receive information wherever it is most convenient at the time.
Our solution is implemented as a series of information collection and retrieval agents:
- Sales Rep Weekly 1:1 Agent
- Lists out scheduled meetings / tasks for each week day
- Proposes additional followups based on open deals
- Dedicates specific prospecting time
- Solicits feedback from sales rep to customise the plan
- Once confirmed, handles calendar + CRM updates for additional tasks / meetings
- End of Day Review Agent
- Uses Weekly 1:1 plan as input for review
- Solicits updates from sales rep on meetings / tasks for the day
- Drafts CRM record and pipeline updates for rep to review
- Once confirmed, updates CRM records on reps behalf
- Sales Manager Digest Agent
- Reviews sales team updates and applies configurable prioritization score (eg. stage, revenue)
- Identifies repeating lowlights from sales team updates for potential training / process improvements (eg customer not available; could not close because x missing)
- Selects team highlights to support weekly sales meeting learnings
- Provides links to specific records referenced for sales manager to review indepth
- Weekly or daily digest cadence depending manager preference
Our agent layer sits between your organization’s communications layer and your systems of record. Updates to any records are tagged as agent generated, clearly differentiating human vs agent updates. To facilitate trust, sales rep updates are not directly reviewable by the sales manager. They are, however, retained in an audit log to support agent QA and root cause analysis where applicable.
How we Measure Success
Our hypothesis for this project is that by reducing data input friction, we can improve both data timeliness and data depth for sales organizations. It rests on the view that manual input is high friction for sales teams to the degree that it drives non-compliance.
Our sales manager briefings build upon improved data ingestion by making it equally accessible, reducing the need for manual manager review. This supports sales managers decision quality by adding nuance and detail to record keeping, while automating key information extraction. Sales managers can act faster on what is important to driving the outcomes they are responsible for.
Specific metrics I would target are:
Development:
- Agent tool use accuracy is ≥ 95%
- ≤ 20% of agent drafted CRM record updates require manual edits
Success Metrics:
- 2x increase in Length in words of record updates (as a proxy for detail)
- Baseline: Current average record update is 12 words.
- 80% of rep activities have CRM records updated within 24hrs (as a proxy for timeliness)
- Baseline: Industry average is 55 hours
Guardrails:
- Agent tool use accuracy must remain ≥ 95% in production. Sustained breach triggers deployment pause pending root cause analysis.