This guide is part of the Claude Cowork for SDRs series — the pillar article covers the complete SDR deployment framework. Here we go deep on the research and scoring workflow specifically: how to configure Cowork for ICP scoring at scale, how to structure multi-source account intelligence, and how to translate that intelligence into personalisation that actually drives replies. For the sequencing side of this workflow, see our guide on writing 10 personalised emails in 20 minutes. For the tool integrations that feed research data into Cowork, see the Outreach and Salesloft integration guide.
Claude Cowork is Anthropic's agentic workspace — it reads files, connects to CRM and data sources via MCP connectors, and executes multi-step reasoning tasks. The research workflow described here uses Cowork's canvas as the central intelligence layer: ICP criteria, prospect data, company context, and competitive positioning all feed into a single session that produces research outputs at a pace manual research can't approach.
Why Manual Research Fails SDRs at Scale
The manual SDR research workflow has a fundamental throughput problem. A properly researched account — company context, buyer profile, pain points relevant to your solution, competitive situation, trigger events — takes 35–50 minutes to produce at quality. At 8 hours per day, that limits a thorough SDR to 10–14 accounts. Sequence writing and admin work reduce actual research output to 6–8 accounts daily.
The result: SDRs constantly choose between research depth and research volume. Go deep and you don't have enough prospects in sequence. Go wide with shallow research and reply rates collapse. Cowork removes this tradeoff — it can match the quality of your best manual research at 7x the output speed, with no degradation across the 50th account of the day.
Reduction in time per account research brief when using Claude Cowork's structured multi-source workflow — from 45 minutes to 7 minutes per account, without sacrificing personalisation quality.
Setting Up Your Cowork Research Canvas
Before you run a single research task, configure your canvas with three foundation files. This setup takes 30–45 minutes once and makes every subsequent research session faster and better calibrated to your market.
ICP-Criteria.md
Define firmographic requirements (company size range, industry verticals, revenue range, geography), technographic signals (existing tools that indicate fit), and hard disqualifiers. Include 3–5 specific trigger events that make a prospect A-tier.
Value-Prop.md
Your solution's 3–4 primary pain points addressed. For each: the problem in the customer's language, the solution mechanism, one concrete outcome with a metric (e.g. "reduces close time from 45 to 12 days for SaaS AEs"), and a customer reference.
Persona-Guides.md
For each of your 2–3 primary buyer personas: their title variants, typical KPIs they're measured on, most common objections, preferred communication style (detailed vs. concise), and the business context that typically drives evaluation of your category.
Battlecards.md
One section per major competitor: their strengths (be honest), their weaknesses, where you win and why, migration trigger scenarios, and how to frame competitive conversations without attacking the competitor directly.
These four files are the persistent intelligence layer of your canvas. Every research brief, ICP score, and personalised message Cowork produces references them. Teams that invest in building thorough canvas foundations consistently outperform those with shallow or missing context files — the quality difference is visible from the first account brief.
ICP Scoring at Scale: The 3-Minute List Qualification Workflow
When you receive a raw prospect list from LinkedIn Sales Navigator, Apollo, or ZoomInfo, the first task is qualification — not research. Running deep research on C-tier accounts is the most common research time drain in SDR workflows. Cowork eliminates this waste by scoring an entire list before you invest any research effort.
Export and clean your raw list
Export your LinkedIn Sales Navigator or Apollo list as CSV. Ensure it includes: company name, company size (headcount), industry, prospect title, and any available technographic or funding data. Remove obvious duplicates. Upload the CSV to your Cowork canvas.
Run the ICP scoring prompt
Reference both the uploaded CSV and your ICP-Criteria.md in the prompt. Ask Cowork to evaluate each row against your criteria and assign tiers. Specify that A-tier requires 4+ core criteria AND at least one trigger signal; B-tier requires 3 criteria; C-tier is anything below.
Review and validate the A-tier list
Scan the scoring output for obvious errors — misclassified industries, outdated headcount data, accounts you already know are dead ends. Cowork's scoring is accurate but CRM data is imperfect. A 5-minute human sanity check on the A-tier list prevents wasted research on the wrong accounts.
Archive B and C tiers appropriately
B-tier accounts go into a "future nurture" Cowork canvas file for lighter-touch sequence later. C-tier accounts are removed unless a specific campaign justifies outreach. This discipline keeps your active research focus exclusively on high-probability accounts.
Multi-Source Account Intelligence: The 7-Minute Research Brief
Once you have your A-tier list, the account brief workflow produces a 300–400 word intelligence document per account that covers everything an SDR needs for call prep, sequence writing, and personalised LinkedIn outreach. The key to speed and quality is running three sources simultaneously.
The three-source approach addresses a fundamental limitation of single-source research: any one source gives you an incomplete picture. The company website shows aspiration. LinkedIn shows current priorities and recent activity. Press releases and news show real strategic events. Cowork synthesises across all three, identifying where they align and where they diverge — the gaps between stated strategy and recent activity are often the richest personalisation angles.
Account Research Brief PromptFor accounts where you don't have pre-gathered source files, Cowork's web search connector can pull company website content and recent news directly into the canvas session. This adds 60–90 seconds per account but eliminates manual source gathering entirely. At scale, the time cost is marginal versus the alternative of manual web browsing for each account.
Deeply researched account briefs an experienced SDR can produce per working day using the Cowork 7-minute research workflow — versus 8–12 briefs per day with manual research.
Translating Research into Personalisation That Drives Replies
Research is only valuable when it translates into outreach that prospects notice. The most common SDR mistake post-research is using account intelligence as decoration rather than substance — mentioning a company's recent funding in the opener but then pivoting to a generic pitch. Cowork's sequence generation, when fed a properly structured account brief, produces copy where the personalisation is load-bearing: the message only makes sense in the context of this specific account's situation.
The principle: every personalised element in your sequence must either explain why you're reaching out now, or explain why your solution is specifically relevant to this account's situation. "Saw you just raised a $20M Series B" is decoration. "Saw you just raised a $20M Series B — expansion capital usually means a 3–6 month sprint to double pipeline before the board expects revenue return, and that's exactly when SDR productivity gaps become visible" is load-bearing personalisation.
For the full sequence writing workflow — how to turn account briefs into 6-touch personalised sequences in under 4 minutes — see our dedicated guide on Claude Cowork for email sequences. For how the research-to-sequence pipeline connects with Outreach and Salesloft for automated delivery, see the sequencing platform integration guide.
Scaling the Cowork Research Operation Across an SDR Team
Individual SDR productivity with Cowork is significant. Team-level deployment multiplies the impact by enabling shared research infrastructure: centralised ICP criteria that update as your ideal customer profile evolves, shared competitor intelligence that improves as deals close, and pooled trigger event intelligence from a monitored account universe that no single SDR could track alone.
The key components of a team-level Cowork research deployment: (1) a master ICP-Criteria.md maintained by the sales manager and synced to each SDR's canvas; (2) a shared battlecard library updated quarterly; (3) a team trigger-event monitor via Cowork Dispatch that covers your full prospect universe and routes alerts to the SDR covering each account; and (4) a weekly "research quality review" where SDRs share briefs that drove strong engagement — feeding successful patterns back into the canvas knowledge base.
If you're deploying Cowork across an SDR team and want a structured implementation, our Claude Cowork deployment service covers the full setup: connector configuration, canvas architecture, team skill library build, and a 30-day onboarding programme. If you're evaluating whether to proceed, book a free strategy call and we'll scope the right setup for your team size and sales motion.
Stop Choosing Between Depth and Volume
Cowork removes the research throughput ceiling for SDRs. Our certified architects configure the full research canvas — ICP files, connectors, team skill libraries — so your team ships production-quality prospect intelligence from day one.