A pipeline plan for the SDR seat at Deepgram

A pipeline plan for the SDR seat

Prepared for the team at Deepgram

By Vincent Hembrick

[email protected] · LinkedIn · (334) 524 1887

About me

Ambitious Scientist

Materials science engineer with patents at J&J MedTech and Aptar. Moved into sales to get closer to the buyer.

Top BDR

Self sourced $6M in pipeline in 15 months at Certara selling into technical buyers within the life sciences industry.

Why Deepgram

My experience building a business as well as my science career give me a unique edge when it comes to selling a highly technical solution.

Opening

Why I built this

To show my approach, excitement, and dedication to the role I built this for the SDR seat at Deepgram specifically. Below is a break down of the ICP, five accounts that match the ICP, example messaging, and a 30, 60, 90 day plan for how I would ramp. Everything is framed against Deepgram's actual GTM and the trigger patterns in the voice AI market right now. If any of it lands, I would value 15 minutes to compare notes.

ICP

Two ICPs, named directly in the JD

The JD names three buyer types: software companies building voice products, co sell partners working with large enterprises, and enterprises solving internal voice AI use cases. Co sell partners cluster with the software builders since both are technical buyers with similar buying behavior, leaving a clean split between voice product builders and enterprise voice operators.

ICP A: Software companies building voice products

Firmographics

Voice AI ISVs, CCaaS and UCaaS, health tech SaaS, conversational AI, AI scribe startups
50 to 5,000 employees, Series A through D
North America heavy, EMEA secondary
Multi provider ASR and TTS stack already in production (Whisper, OpenAI Realtime, Google STT, ElevenLabs)

Buyer titles

CTO and Co Founder & CTO
VP Engineering, Head of Speech AI
Head of AI, Head of Conversational AI
Founder and Founding CEO at pre Series B
Chief Product Officer at established voice platforms

Pains and triggers

Hyperscaler ASR billing crossing $1M annually with quarterly cloud cost reviews now visible at the CFO level
Enterprise sales blocked from healthcare, finance, and government deals because consumer cloud ASR cannot pass HIPAA, FedRAMP, or PCI
Latency variance from existing providers showing up in production NPS and customer escalations
Concurrent stream limits during peak events (Black Friday, AEP, election cycles)
Series B or C close that frees infrastructure budget for vendor consolidation
Multi provider integration debt eating engineering velocity that the team would rather spend shipping product

ICP B: Enterprises with internal voice AI use cases

Firmographics

QSR chains (50 plus locations), Tier 1 BPOs, telcos, large insurance carriers, IDN healthcare
5,000 plus employees, public markets or PE backed
North America primary, EMEA enterprises secondary
Already running voice in production (call center, drive thru, IVR, claims)

Buyer titles

Chief Digital Officer, Chief Information Officer
Chief Transformation Officer, Chief Operating Officer
VP Operations, VP Customer Experience
VP IT, VP Restaurant Technology

Pains and triggers

Failed incumbent voice AI deployment (the McDonald's and IBM exit is the canonical 2024 case study) creating board level redeployment pressure
Hyperscaler ASR contract overruns at scale, opaque billing, vendor lock in pressure
Labor as percent of revenue at 28 to 32 percent versus 22 percent target, CFO pressure mounting
Peak event capacity crisis (AEP for health insurance, Black Friday for retail, IRROPS for airlines)
Client AI mandates pushing BPOs to absorb GenAI cost reduction within 12 to 18 months
New vertical expansion or M&A integration creating new compliance footprint

Firmographic ranges and buyer titles for ICP A are inferred from Deepgram's public customer footprint, the SDR JD's three buyer types, and the voice AI market analysis I worked through. ICP B verticals are pulled from the Deepgram for Restaurants page, partnership pages, and industry analyst coverage. Worth verifying against the live ICP doc before working.

Targets

Five accounts I would stack rank day one

Best fits across both ICPs, no forced 3+2 split. I checked each one against Deepgram's public customer wall, partnership pages, and recent press before listing.

Heidi Health (Melbourne and San Francisco). Series B AI medical scribe expanding into US clinical pilots, raised $16.6M in 2024. Microsoft DAX is the entrenched competitor. ASR cost per encounter is the unit economics gating their pre Series C planning. ICP A.

Sierra (San Francisco). Bret Taylor's AI agent company at $4.5B valuation, voice and chat agents for enterprise customer service. Sells into financial services and retail enterprises that demand SOC 2, HIPAA, and audit grade infrastructure. ICP A.

Cresta (Palo Alto). Series C conversational AI for contact centers with a Twilio Flex partnership. Sells to BPOs and large insurance carriers running multi language at scale. Their enterprise compliance story is gated by ASR provider deployment options. ICP A.

Restaurant Brands International (Toronto). Parent of Burger King, Tim Hortons, Popeyes, and Firehouse Subs with roughly 30,000 global locations. Public CFO commentary on labor cost discipline and drive thru voice AI pilots reported at Burger King in 2024. The McDonald's IBM exit is the redeployment template. ICP B.

Concentrix (Newark, CA). Tier 1 BPO post Webhelp merger, roughly 440,000 employees and $9B plus revenue. Public iX AI first transformation strategy with explicit client mandates to absorb GenAI cost reduction. Multi language and multi region deployment requirements push toward self hosted ASR. ICP B.

All five are illustrative best fits. I verified each one is not a current customer of the hiring company (homepage, case studies, recent press) before listing. I would still dedupe against the existing CRM and pull in real time triggers (job postings, funding, M&A, breach proximity, partner announcements) before adding any to a sequence.

Cadence

A four touch sequence in Deepgram's voice

Persona: CTO and co founder at a Series B AI medical scribe, 50 to 200 employees, US and AU footprint. Example target: Heidi Health, illustrative only. Cadence: 4 touches over 12 days, mixing email, LinkedIn, and phone. Voice calibrated to Deepgram's plain spoken, benchmark backed tone.

Touch 1: Day 1, Email pattern interrupt (T.I.P.S.)

Subject: Medical ASR

Hey Stefan, Saw the US clinical pilots ramping up and the Series B announcement. Imagine ASR cost per encounter and BAA execution speed are both at the top of the priority stack right now, especially with Microsoft DAX moving aggressively into the same provider conversations you are running. Most AI scribe teams I track hit a wall when medical WER stalls under 92 percent on drug names and clinician trust drops the same week. The teams that push past that wall tend to share two things: voice native models trained on real clinical audio, and deployment that lets them run inside the customer VPC without a six month security review. Two scribe platforms cut their ASR spend roughly 50 percent moving from hyperscaler STT to Nova-3. Self hosted, BAA same week, roughly a third of Azure at scale. Mind if I share the medical WER benchmark? Vincent

Touch 2: Day 3, LinkedIn connection (Relevant Message 2)

Hey Stefan, dropping you a connect. Saw the US clinical pilots ramping and the Microsoft DAX comparisons in your investor materials. Mind if I ask a few questions about how you are managing ASR cost per encounter as encounter volume scales? P.S. The integration with the Aussie practice management systems is a clever wedge. Have not seen anyone else solve that cleanly.

Touch 3: Day 6, Cold call with voicemail (Direct to email, structure 3)

30 to 40 seconds when read aloud

Hey Stefan, it's Vincent. I noticed Heidi is in the middle of US clinical pilots while Microsoft DAX still dominates the AI scribe space. Most teams in your spot are wrestling with how to keep ASR cost per encounter sub dollar while improving medical WER. I emailed you a few ideas. Subject line is Medical ASR. Hope they help. Speak soon.

Touch 4: Day 12, Break up email (Thoughtful Bump with case study)

Subject: Medical Benchmark

Hey Stefan, Given ASR cost per encounter and medical WER were both priorities for the US clinical pilots, thought this would be worth flagging before I close the loop. Quick breakdown on how a Series B scribe platform cut WER on drug names by 7 points and pulled ASR spend from $1.40 to $0.42 per encounter after moving to Deepgram. Same kind of clinical sample sets you would test against, similar encounter volumes, similar BAA timeline pressure from health system buyers. The piece I keep coming back to is the deployment side. Most teams I track lock in self hosted Nova-3 because the alternative is a six month InfoSec review every time a new health system asks where audio data sits at rest. Voice native model, runs in your VPC, BAA same week. Worth a peek? [link to case study] Vincent

Plan

How I would ramp at Deepgram in my first 90 days

The plan below mirrors how I ramped at Certara, where I went from 0 sales experience to top of the BDR table by pairing tight learning loops in month one with disciplined high volume execution through months two and three.

DAYS 1 TO 30

Learn the voice AI stack and Deepgram's edge

Articulate what Deepgram solves at the CTO, VP Product, and CEO level for all three buyer types named in the JD: software builders, co sell partners, and enterprises
Shadow top SDRs and AEs across each motion to see how each buyer type gets worked
Listen to recorded discovery calls and demos to internalize how Nova-3, the Voice Agent API, and self hosted deployment land in technical conversations
Get fluent in Salesforce, Outreach, Sales Navigator, Apollo, and ZoomInfo
Study three competitors (Google STT, AssemblyAI, and Microsoft DAX for healthcare) and map where Deepgram wins on accuracy, latency, deployment, and pricing
By day 30, have my own call script and personal email playbook tied to the three highest leverage trigger types: hyperscaler bill review, compliance blocked deals, and post McDonald's IBM redeployment

DAYS 30 TO 60

Run the cadence, hit ramp quota

Lock the daily schedule: research and list building 9 to 10, cold calls 10 to noon, email and LinkedIn outreach 1 to 3, follow ups and CRM hygiene 3 to 5
Build a 100 account named list weighted across both ICPs and run the 4 touch sequence at full volume
Hit the Month 2 ramp quota
Roleplay technical objections daily with peer SDRs and a Sales Engineer: latency benchmarks against hyperscalers, BAA execution speed for healthcare buyers, FedRAMP positioning for federal opportunities, and the McDonald's IBM redeployment story for QSR
Weekly feedback loop with my manager on what is converting, then iterate the framework week by week
Fail fast, log everything, adjust

DAYS 60 TO 90

Analyze the funnel, build the personal playbook

Pull conversion data across the full funnel: dials to connects, connects to meetings, meetings to SQLs, broken out by ICP and by trigger type
Identify which trigger and persona combinations produce the strongest pipeline contribution and double down
Consistently hit quota and lead the SDR team in call activity
Continue ongoing development through voice AI podcasts, the Deepgram blog, and weekly product updates so I stay sharp on a category that moves every month
By day 90, have a personal playbook documenting which triggers, personas, and openers work best for me at Deepgram, ready to share back to the team
That playbook is the artifact that signals AE track readiness, the 18 month goal I would be working toward

Closing

If any of this is in the ballpark of how the Deepgram team is thinking, I would value a 15 minute conversation to compare notes on what is working, what is not, and where I would plug in fastest. If the plan above misses the mark, that is also useful feedback for me. Either way, thank you for reading.

Vincent Hembrick

[email protected] · LinkedIn · (334) 524 1887