Join the Waitlist

Use CasesFinancial services

Meridian Capital Group

Find the line. Hold it.

A capital markets group that builds and operates the settlement, treasury and reporting infrastructure its clients used to build badly in-house.

workflows shown here
5
AI agents behind them
6
runs seeded in the tenant
13
people across 4 Dave roles
5

Compliance is the gate on everything Meridian publishes or opens. Dave runs the screening, the scanning and the drafting, and routes only the risky items to a person.

Dave tenant
demo-finserv
Workspace
Meridian Workflow Hub
Sector
Financial services and capital markets technology
Who reads their work
Two audiences on one brand: institutional buyers reading market commentary, and the client-side engineers reading integration runbooks.
  • Exact
  • Fast
  • Legible
  • Unshowy

Can AI do KYC and AML pre-screening?

It can do the reading. It cannot do the deciding, at least not here. In this demonstration tenant the KYC Pre-Screen Analyst takes the client name, account type and jurisdiction, screens them against KYC and AML risk indicators, and writes a risk rating of low, medium or high. Its prompt tells it to recommend only, never to approve or reject. The Compliance Approval step is the only thing that can clear an account, and a rejection returns the run to the intake form.

Can a workflow call an outside system while it is running?

Yes, and two of Meridian's do. The SEC Filing Analysis workflow makes a live call to the SEC's public EDGAR submissions endpoint with a declared User-Agent, then maps the form types and report dates it gets back into the run's own fields for the agent to write from. The Website Builder workflow makes five such calls in a row: create a repository, commit the site, look up the subdomain, deploy the worker and turn the route on.

The workflows, node by node

Not a summary of them. These are the graphs as they are seeded: every step in order, its type, the agent on it, who it is assigned to, and where a rejection sends the run back to.

Guide: Build your first workflow

Client Onboarding Compliance Review

A KYC and AML pre-screen on every new account. The agent produces a structured risk view; a compliance officer is the only one who can clear the account.

5 nodes

  1. Start

    Start

  2. User Interaction

    Gather Client Details

    Submitter

    Client name, account type and jurisdiction, all three required before the run can move.

  3. Agent Interaction

    AI KYC Pre-Screen

    KYC Pre-Screen Analyst

    Writes a risk rating of low, medium or high, with two retries before a person is pulled in.

  4. Human Review

    Compliance Approval

    Reviewer

    Sees the rating the agent proposed and can add review notes. Only this step clears the account.

    Reject returns to Gather Client Details

  5. End

    Account Cleared

Marketing Material Compliance Check

Every outbound piece is scanned for regulatory red flags. Low and medium risk clears itself. High risk cannot: it is routed to a human compliance reviewer by rule, not by judgement.

6 nodes

  1. Start

    Start

  2. User Interaction

    Submit Material

    Submitter

    Title and body of the material going out.

  3. Agent Interaction

    AI Compliance Scan

    Marketing Compliance Scanner

    Flags promissory language, unbalanced performance claims, missing disclosures and testimonial issues, each with a severity.

  4. Routing

    Risk Triage

    High-risk findings require human compliance review; low and medium auto-clear.

    • High RiskCompliance Review
    • ClearedMaterial Cleared

  5. Human Review

    Compliance Review

    Reviewer

    Reached only on the high-risk branch.

    Reject returns to Submit Material

  6. End

    Material Cleared

SEC Filing Analysis

The one that leaves the building. An API Call node hits the SEC's public EDGAR submissions endpoint for a company's filing index, then an agent turns the index into a research brief for an analyst to check.

6 nodes

  1. Start

    Start

  2. User Interaction

    Set Target Company

    Submitter

    Company name and its ten-digit zero-padded CIK from EDGAR company search.

  3. API Call

    Fetch EDGAR Filing Index

    A live GET to data.sec.gov with a declared User-Agent, per the SEC's fair-access etiquette. Form types and report dates are mapped back onto the run's fields.

  4. Agent Interaction

    AI Research Brief

    Filing Research Analyst

    Told in its prompt that it sees index metadata only, never full filings, and must direct the analyst to the source document rather than invent a figure.

  5. Human Review

    Analyst Review

    Reviewer

    Rejection sends it back to the agent, not back to the analyst who started it.

    Reject returns to AI Research Brief

  6. End

    Brief Filed

Quarterly Risk Report Assembly

The agent assembles the quarterly risk narrative and emits it as an HTML file. The risk officer signs it off, or sends it back to be redrafted.

5 nodes

  1. Start

    Start

  2. User Interaction

    Set Reporting Scope

    Submitter

    Which quarter.

  3. Agent Interaction

    AI Draft Report

    Risk Report Drafter

    Executive summary, key risk indicators, changes against the prior quarter and recommended actions, written out as an HTML artifact.

  4. Human Review

    Risk Officer Sign-off

    Reviewer

    Reject returns to AI Draft Report

  5. End

    Report Published

Website Builder

Twelve nodes, five of them calls to external services. A brief goes in, an agent generates the page and its worker script, the workflow creates a repository, commits the site, deploys it and turns the route on, then a person reviews what actually went live before an agent gives the final go or no-go.

12 nodes

  1. Start

    Start

  2. User Interaction

    Gather Site Brief

    Requester

    Site name, brief, style notes, plus the repository, worker and account identifiers the deploy steps need.

  3. Agent Interaction

    AI Generate Site

    Website Generator

    Produces a complete self-contained page and the worker script that serves it.

  4. API Call

    Create GitHub Repo

  5. API Call

    Commit Site to Repo

  6. API Call

    Get workers.dev Subdomain

  7. API Call

    Deploy Cloudflare Worker

  8. API Call

    Enable workers.dev Route

  9. Agent Interaction

    Compose Preview URL

    Website Generator

    Assembles the preview address from the fields the deploy steps wrote back.

  10. Human Review

    Review Live Site

    Reviewer

    A person looks at the deployed page. Rejecting sends it back to generation, not to the brief.

    Reject returns to AI Generate Site

  11. Agent Review

    AI Go/No-Go

    Website Generator

    Reject returns to AI Generate Site

  12. End

    Site Live

The agents, and what each one is forbidden to do

An agent in Dave is a configured worker, not a chat box: a written prompt, a pinned provider and model, and its own context. The line that matters most in each of these prompts is the constraint.

Guide: Create an agent

KYC Pre-Screen Analyst

Screens new client details against KYC and AML risk indicators and produces a structured pre-screen summary.

GuardrailNever approve or reject. Recommend only.

Marketing Compliance Scanner

Reviews marketing material for regulatory red flags: performance claims, guarantees, missing disclosures.

GuardrailOutput findings with a severity of info, warn or high, and a suggested rewording.

Risk Report Drafter

Assembles quarterly risk report narratives from portfolio and incident data.

GuardrailProfessional, neutral tone. No recommendation the data does not support.

Filing Research Analyst

Summarizes a company's recent SEC filings into a structured research brief for investment review.

GuardrailIt sees index metadata, not full filings. It must never invent a financial figure, and must point the analyst at the specific filing instead.

Website Generator

Generates a complete single-page website and its worker wrapper from a creative brief.

GuardrailSelf-contained output only: inline styles, responsive, no external assets.

Expense Policy Checker

Validates expense submissions against company policy: limits, categories, receipts and justification quality.

GuardrailAssess against policy. It does not approve or reject; the routing rule and the manager do that.

Who is in the workspace

Five seats, and the Dave role each one carries. Role decides what a person can reach; the workflow decides what lands in their inbox.

  • Head of ComplianceAdmin
  • Workflow DesignerCurate
  • Client Onboarding AnalystUse
  • Portfolio ManagerUse
  • Risk OfficerReporting

What is in flight

The runs seeded into this tenant, and for each one still moving, the step it is waiting on.

Guide: Watch a run and read its timeline
  • RunningOnboarding: Atlas Verde Holdings S.A.Waiting on Compliance Approval
  • RunningLinkedIn Campaign: Direct Indexing LaunchWaiting on Compliance Review
  • RunningQ3 2026 Enterprise Risk ReportWaiting on Set Reporting Scope
  • CompletedOnboarding: Whitfield Family TrustApproved, rated low
  • CompletedQ3 Wealth Insights NewsletterApproved after edits
  • CompletedFiling Review: Apple Inc. (AAPL)Brief accepted

What one of these agents actually produced

An excerpt from a real run in this tenant. Read what it does and does not do: it grades every finding and hands the decision to the next node.

compliance-scan-linkedin-campaign.txtAI Compliance Scan, by the Marketing Compliance Scanner
COMPLIANCE SCAN FINDINGS: Direct Indexing LinkedIn Campaign [HIGH] Post 2: "save thousands in taxes every year" is an unsubstantiated quantified benefit; requires substantiation or removal.[HIGH] Post 5: hypothetical tax-loss harvesting example lacks required assumptions disclosure.[WARN] Post 6: CTA implies personalized advice without suitability language. OVERALL RISK RATING: HIGH. Routed to human compliance review.

The rule does the routing, not the reviewer. Two of these five workflows send work to a person only when a field crosses a line the firm set in advance, which is what stops a compliance queue filling up with things that never needed a human.

Every fact on this page is read from this tenant's seed data (packages/backend/prisma/demo-packs/financial-services.ts, with templates/sec-edgar.ts, templates/website-builder.ts and templates/expense-report.ts, demo-packs as of 2026-08-29) or from the firm's own brand pack. Nothing here is a customer story.

The other two

Northwind TherapeuticsPharmaceuticalHartwell & Associates LLPLegal

Frequently asked questions

Is Meridian Capital Group a real company?

No. Meridian Capital Group is a fictional firm we invented as a demonstration tenant. It is not a customer of HelloDave.ai, it has no website, and this page is not a case study. What is real is the tenant behind it: every workflow, node, routing rule, agent guardrail and run described here is read out of the seeded demo-finserv Dave tenant.

What happens when a reviewer rejects a run?

A rejection carries a written reason and returns the run to a node the workflow author chose in advance, not simply to the previous step. On the onboarding review it goes back to the intake form so the details can be corrected. On the SEC filing brief it goes back to the agent to be rewritten, rather than back to the analyst who started it.

How does a compliance scan avoid sending everything to a human?

A Routing node reads what the AI compliance scan found. High-risk material requires human compliance review; low and medium risk clear themselves. The rule is written into the workflow in advance, so what reaches the compliance queue does not depend on who happens to be looking at it.

Now build yours.

Same canvas, same node types, your process. Put the first version on it and watch one run end to end.

Join the WaitlistAll three use cases
Financial Services Use Case: KYC and Compliance Workflows | Dave