# Account Intelligence: Configuration Guide > **Who this is for:** Customers who have been provisioned for Account Intelligence and are ready to configure their workspace. > **Goal:** Complete your Account Intelligence setup — including integrations, scoring, personas, research agents, and user roles — so your team can start using it right away. *** ## Before You Begin Before diving into configuration, make sure the following are in place: 1. Confirm that **Account Intelligence** is visible in your HockeyStack sidebar. 2. Open **Atlas** from the sidebar. If you don't see it, contact your HockeyStack admin to grant you workspace admin access. 3. In Atlas, check the **Integrations** section and complete any that are still marked incomplete. 4. Confirm that at least one **Funnel** with stages exists under **Funnel Stages**. If not, create one before continuing. 5. If you have Salesforce CRM, confirm that your **Salesforce iFrame** is connected and enabled. Once all of the above are in place, you're ready to configure Account Intelligence. *** ## Overview {% embed url="" %} Note: your instance may look different from what's shown in the video, depending on your access level and plan ## Step 1: Add Users and Roles Before configuring any account intelligence features, set up your team so the right people have access. Users and roles are managed in **Workspace Settings → Team**. Only **Admin** users can access this section.
### 1a. Invite Team Members 1. Navigate to **Workspace Settings → Team**. 2. Click **Invite Member**. 3. Enter the user's email address. 4. Select an **access level**: * **Admin:** Full access to platform, configuration, and workspace settings. Recommended for ops admins. * **Member:** Full access to platform and configuration. Recommended for Team-Leads and Managers * **Viewer:** Full access to platform, no access to configuration. Recommended for IC’s. 5. Under **Products**, grant access to **Sales**. 6. Click **Invite**. The user will receive an email and appear in the team list as "Invited" until they accept.
### 1b. Create User Roles User roles are separate from access levels. They define categories within your sales team (e.g., AE, SDR) and are required for users to be assigned to workflows. 1. In **Workspace Settings → Team**, open the **User Roles** section. 2. Click **Add Role**. 3. Enter a **name** that reflects the function (e.g., "AE", "SDR", "Sales Manager"). 4. Click **Save**. 5. Repeat for each role your team uses.
### 1c. Assign Roles to Team Members 1. In the team member list, click **Edit** on a member's row. 2. Set their **User Role** from the dropdown (e.g., AE or SDR). 3. Optionally, set the **Reports To** field to establish a reporting hierarchy, this allows managers to view their team's activity. 4. Click **Save**. > 💡 **Tip:** Every rep who will use Account Intelligence should have a role assigned. Users without a role assigned will have limited functionality. *** ## Step 2: Create Target Account Views Account views define which accounts are scored and researched by Account Intelligence. Everything downstream; scoring, stakeholder discovery, AI research - is scoped to the views you configure here. 1. In the sidebar, navigate to **Companies → All Companies**. 2. Open the **view selector** dropdown in the toolbar and click **Create new view**. 3. Give the view a clear, descriptive name (e.g., *Tier 1 Target Accounts - Enterprise*). 4. Apply filters to define which companies appear in this view. Useful filter types include: * **Company properties:** Industry, employee count, annual revenue, country * **Owner:** Assign views to teams of reps. (Note: Do not create views that are specific to 1 rep - for scoring, stakeholder discovery, and AI research, it is better to run these on team-wide views.) 5. Aim to keep each view under **5,000 accounts** for best performance. 6. Click **Save for everyone** to make the view available to your full team. > ⚠️ **Important:** Views are shared workspace-wide. When you save a view, all team members will see it. Use descriptive names so it's clear what each view represents. > 💡 **Tip:** Start with one view that captures your core target accounts for the whole team. You can create additional views later to segment by tier, region, or team ownership.
*** ## Step 3: Configure Account Scoring Account Scoring determines how HockeyStack evaluates and prioritizes your accounts. Scores are computed automatically and appear on each company page to help your team focus on the accounts most likely to convert. **Atlas > Activate > Scoring** #### 3a. Set Up Account Scoring Account Scoring uses a machine learning model that learns from your historical data to predict how likely each account is to reach a goal — such as closing a deal or booking a demo. Instead of manually weighting signals, the model studies the patterns shared by accounts that have already converted and scores every account from **0 to 100** based on how closely it matches those patterns. Scores update automatically and come with **scoring factors** that explain *why* each account scored the way it did. Scores are configured from **Atlas → Scoring**. You can create multiple scores, each tied to a different conversion goal (e.g., one for "Closed Won" and another for "Demo Booked"). **To create a score**, click **Create Score** and complete the guided wizard: 1. **Name your score** — Use a descriptive name your team will recognize (e.g., "Deal Won Score", "Demo Readiness Score"). 2. **Choose the score type** — Select **Behavior Score** to score on engagement signals like page visits, content downloads, email opens, and event attendance. *(Fit Score, based on firmographic data, is coming soon.)* 3. **Choose the entity level** — Select **Company Score** to aggregate all activity across an organization into a single score. *(Person Score is coming soon.)* 4. **Set a conversion goal** — Pick the outcome the model should learn to predict (e.g., "Deal Won", "Demo Booked", "MQL Achieved"). The model studies accounts that have already reached this goal to learn what they have in common. 5. **Select your account list(s)** — Choose one or more **Sales Views** that define which accounts to score (use the view(s) created in Step 2). For best results, pick a broad view such as *All companies with domain* so the model has enough training data. 6. **Scoring details (Behavior Scores)** — Optionally add up to **3 Touchpoint Properties** (engagement signals like *Channel* or *Asset Type*) and up to **3 Score Breakdowns** (dimensions like industry or company size that train a separate model per segment for more precise scoring). 7. **Field name for CRM sync** — Provide the custom field/property name to use when syncing this score to Salesforce or HubSpot. Confirm field visibility settings with your CRM admins. 8. **Use in Blueprints** — Toggle on to power the Marketing Blueprints feature with this score's data. Click **Save**. The model begins training and shows a **"Running"** status; once it changes to **"Completed"** (usually a few minutes, depending on data volume), scores are available across the platform — on the Companies list, on individual company pages, and synced to CRM. **To edit or delete a score**, use the Scoring Configuration screen. Each score in the list shows its name, type (Behavior or Fit), entity level (Company or Person), conversion goal, and current processing status. > 💡 **Tip:** Start with your primary conversion goal (e.g., "Closed Won") and validate it against recently closed deals — would the model have scored them highly? — before rolling out additional scores. The properties most teams find useful as Touchpoint Properties are *Channel* and *Asset Type*. > ⚠️ **Note:** Accounts with a Closed-Won deal (already converted) or no verified domain are not scored. Scores recalculate automatically with a **recency decay** that gradually lowers the score of accounts that go quiet, so stale engagement doesn't keep an account artificially high. Do **not** manually trigger a rescore — reach out to your HockeyStack contact if you believe one is needed. More details available [here](/account-intelligence/account-scoring.md). *** ## Step 4: Define Personas Personas define the buyer profiles your team targets within prospect organizations. They drive stakeholder discovery, contact identification, and how the AI researches and summarizes accounts. Without well-defined personas, the system has no targeting criteria. 1. Open **Atlas → Personas**. 2. Click **Add Persona**. 3. Fill in the required fields: * **Name:** A unique, descriptive label (e.g., "VP of Engineering", "Security Decision Maker"). * **Department:** The organizational function this persona belongs to (e.g., Engineering and Technical, Sales, Finance). * **Job Levels:** Select one or more seniority tiers that apply (e.g., C-Level, Director, Manager). At least one is required. * **Description:** A natural-language explanation of who this persona represents. Maximum 1,000 characters. See guidance below. 4. Optionally fill in: * **Subdepartments:** Narrow the department further (e.g., under Engineering: "Cybersecurity", "DevOps", "Cloud / Mobility"). * **Job Title Keywords (Include):** Keywords that must appear in a contact's title to match (e.g., "security", "infrastructure", "platform"). Comma-separated. Use this when department + level alone is too broad. * **Job Title Keywords (Exclude):** Keywords that disqualify a contact even if other criteria match (e.g., "intern", "assistant", "coordinator"). Use this to filter out junior or adjacent roles. 5. Click **Add Persona** to save. 6. Repeat for each buyer persona relevant to your sales process. ### Writing Strong Persona Descriptions The description is the most impactful input for AI quality. A vague description produces vague outputs. Include: * **What they're responsible for:** "Owns the cloud infrastructure budget and vendor relationships" * **What they evaluate:** "Evaluates observability platforms, APM tools, and cost optimization solutions" * **Who they report to:** "Reports to CTO or VP of Engineering" * **Key pain points:** "Concerned with mean time to recovery, deployment frequency, and cloud spend" * **Why they matter:** "Often the technical champion who drives internal evaluation" **Example of a strong persona: "DevOps Decision Maker":** * Department: Engineering and Technical * Subdepartments: DevOps, Cloud / Mobility * Job Levels: Director, Head, Manager * Include Keywords: devops, infrastructure, platform, SRE, reliability * Exclude Keywords: intern, junior, associate * Description: *"Technical leaders responsible for infrastructure and deployment strategy. They evaluate CI/CD tools, cloud platforms, and developer productivity solutions. Often report to CTO or VP of Engineering. Key pain points include deployment velocity, incident response, and infrastructure costs."* **Example of a poor persona: "Engineering":** * Department: Engineering and Technical * Job Levels: Director, Head, Manager, Senior, Specialist * Description: *"People in engineering."* This fails because it's too broad and gives the AI nothing to work with. > 💡 **Rule of thumb:** If your persona would match tens of people at a mid-sized company or hundreds at a large one, it's too broad. Narrow it down. > 💡 **Tip:** For Job Title inclusions, try not to add entire job titles (VP of engineering) try to include individual keywords (devops) - additionally, it is best practice to start with only adding Job Title exclusions first, then adding inclusions if the people sourced are still too broad. ### How Many Personas to Create Start with 2–3 well-defined personas representing your primary buyer and technical champion. A few precise personas will outperform many vague ones. Add personas for economic buyers, influencers, or blockers as your process matures. ### Testing Your Personas Before enabling personas across all your accounts, validate them on a single company: 1. Open a persona for editing. 2. In the right panel, find the **Test Stakeholder Discovery** section. 3. Enter a **Company Domain** (e.g., `acme.com`). 4. Click **Run** and review the discovered contacts. 5. Verify that the people found match your expectations, both that the right people are included and that irrelevant people are not. 6. If results look off, adjust the department, job levels, or keywords and test again. *** ## Step 5: Enable Stakeholder Mapping Stakeholder Mapping automatically identifies and maps key contacts at your target accounts based on your personas. Discovered contacts appear in each company's **People** tab with email addresses and LinkedIn profiles. 1. Navigate to **Atlas → Personas** and switch to the **Stakeholder Mapping** tab. 2. Click **Add View** to select which account views to run discovery across (use the view(s) created in Step 2). 3. Review the mapping table, which shows: * **Companies:** Total accounts in the view * **With Stakeholders:** Accounts with at least one discovered contact * **Stakeholders:** Total contacts discovered * **With Email:** Contacts with a verified email address 4. Click **Run Now** on the view row to trigger discovery immediately. Discovery will also run automatically every 7 days going forward, and overnight after any persona changes. > ⚠️ **Note:** Views with more than 15,000 accounts are not supported for stakeholder mapping. If your view exceeds this limit, create a more targeted filtered view before enabling mapping. *** ## Step 6: Set Up AI Research Properties AI Properties are custom research signals the AI evaluates for each company in your target views. You define the question, the system researches every account to produce a structured, evidence-backed answer. These signals appear on company pages, help prioritize accounts, and inform how the AI reasons about each account.
### Create an AI Property 1. Navigate to **Atlas → Research Agents**. 2. Click **New Property**. 3. Fill in: * **Name:** A clear label for what you're researching (e.g., "Recent Funding Round", "Hiring for Data Roles"). * **Instructions:** Tell the AI exactly what to look for and how to evaluate the answer. Be specific (see guidance below). * **Output type:** Choose the format that fits your signal: * **Yes / No:** Binary answer (e.g., "Does this company use Kubernetes?") * **Dropdown:** One value from a list you define (e.g., industry vertical) * **Score:** A 0–10 numeric rating (e.g., "How strong is their product-market fit signal?") * **Freeform:** Open-ended text (e.g., "What is their primary competitive differentiator?") * **Refresh Every (days):** How often the answer should be re-evaluated (e.g., 7 for fast-changing signals, 30 for stable ones). 4. Click **Save**. ### Writing Effective Instructions Vague instructions produce vague answers. Include: * **Instructions:** “Your goal is to identify whether this company has had a recent funding round in the last 90 days” * **What Sources to use:** "Check the company's careers page, recent press releases, and LinkedIn job postings for roles mentioning Kubernetes, Docker, or container orchestration." * **How to evaluate:** "Answer Yes only if there is direct evidence of production usage, not just job postings mentioning the technology." * **Edge cases:** "If the company is a consultancy that implements the technology for clients but does not use it internally, answer No." > 💡 **Tip:** If you’re having a hard time creating good research prompts, then you can leverage Claude or ChatGPT for prompt building assistance. ### Enable Auto Research Once a property is created, enable auto research so it runs on a recurring schedule across your target views: 1. Open the property in the **AI Agent Builder**. 2. In the **Auto Research** section, toggle **Enable Auto Research**. 3. Confirm the **Refresh Every (days)** interval. 4. Select the **account view(s)** to target. 5. Click **Save**. > 💡 **Tip:** Start with 3–5 high-value signals rather than trying to track everything at once. Good signals are things your team would research manually if they had unlimited time; recent funding, specific hiring activity, technology usage, or expansion into new markets. *** ## Step 7: Add Products Products give the AI context about what your team is selling. This context shapes how accounts are researched and summarized. 1. Navigate to **Atlas → Products**. 2. Click **Add Product**. 3. Fill in the following fields: * **Name** (required): The product name as your team refers to it (e.g., "Revenue Intelligence Platform"). * **Description** (recommended): A detailed explanation of what the product does, who it is for, and its key value propositions. See guidance below. 4. Click **Save**. 5. Repeat for each distinct product or solution your team sells. ### Writing Effective Product Descriptions The description is read by the AI when generating account insights. Include: * **What it does:** e.g., "An AI-powered platform that surfaces buying signals and maps stakeholders at target accounts." * **Who it's for:** e.g., "Built for B2B sales teams with 10+ reps selling into enterprise accounts." * **Key differentiators:** e.g., "Unlike generic tools, insights are grounded in the customer's own CRM and engagement data." > 💡 **Tip:** If your company sells multiple products but one is almost always the entry point, focus your primary product description on that product. Describe the others as expansion or add-on offerings, and note how your team typically sequences them. *** ## Step 8: Add Business Context Help Account Intelligence understand your business so it can deliver more relevant insights across all accounts. 1. In **Atlas**, navigate to the **Business Context** section. 2. Upload any relevant context documents. For example, your ICP definition, competitive positioning, or company overview. 3. Add a high-level description of your company and what you do. This context is used by the AI when generating account summaries, scoring reasoning, and research outputs. *** ## Step 9: QA Your Account Details Before rolling out to your full team, review a sample of accounts to verify everything is configured correctly. 1. Open your **Target Account View** and click into several accounts. 2. For each account, review the following tabs: * **Overview:** Does the account summary look accurate and relevant? * **Journey:** Are funnel stages and touchpoints appearing correctly? * **People:** Are the right contacts being surfaced based on your personas? * **Stakeholders:** Are personas mapping to the right individuals? * **Signals:** Are your AI Properties returning relevant, useful results? 3. Review **at least 3 accounts**, and do one complete, thorough pass on a single account. 4. If anything looks off, return to the relevant configuration step and adjust. *** ## Step 10: Set Up Workflows *(Optional)* Workflows allow you to automate actions based on Account Intelligence data — for example, pushing signals or scores into Salesforce fields, or triggering alerts when an account's status changes. This step is optional and depends on your team's specific use case. If you'd like to configure workflows, reach out to your HockeyStack contact for guidance on the best setup for your situation. *** ## You're Ready 🎉 Once you've completed the steps above, your team is set up to use Account Intelligence. Here's a quick summary of what you've configured: * ✅ Team members invited and roles assigned * ✅ Target account views created * ✅ Account scoring configured * ✅ Personas defined and tested * ✅ Stakeholder mapping enabled * ✅ AI research properties live * ✅ Products defined with clear descriptions * ✅ Business context added * ✅ Account details reviewed and validated *** ## Need Help? If you run into any issues during setup, reach out to your HockeyStack onboarding contact or visit our documentation at [agents-docs.hockeystack.com](http://agents-docs.hockeystack.com). --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.hockeystack.com/account-intelligence/account-intelligence-configuration-guide.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.