Case Study: How an Architecture Consultancy Saved VND 360 Million a Year with an AI Agent for Meeting Data Security

1. Pain: The meeting ends, but the risk begins

Minh, the director of an architecture consultancy in Ho Chi Minh City, handled hundreds of recordings, transcripts, and meeting minutes with clients and contractors every year. The data was scattered across meeting platforms, email, shared folders, and personal computers.

One public link or excessive permission could expose bid prices, exclusive designs, customer information, or contract terms. The company had no automatic way to identify sensitive content or detect bulk downloads and sharing with unrelated external parties.

Managers were still running operations manually: searching for minutes, reminding owners, and reviewing permissions file by file. This created technical debt. Every quick fix added another layer of patchwork and made control more difficult.

2. Agitate: Thirty careless minutes can cost hundreds of millions

Before automation, Minh's operations team spent around 30–40 hours every month compiling minutes, locating decisions, and checking access rights. This work did not generate new revenue; it merely kept the process from going down.

The larger loss was opportunity cost. Skilled architects were pulled into administration, project schedules slowed, and managers tracked vanity KPIs instead of removing bottlenecks. If a design or bidding strategy leaked, the company could lose a contract, pay for crisis handling, and damage its reputation with the client.

Using shared folders as a quick fix may look efficient, but it is only half-baked optimization. After accounting for staff time, mistakes, and accidental sharing, the company estimated its annual exposure at approximately VND 360 million.

3. Solve: Deploy a controlled AI Agent in 3 steps

Step 1 – Standardize data and classify risk. HimiTek connects an AI Agent to the company's existing workflow to receive transcripts, minutes, and attachments. The Agent labels content as internal, partner-facing, or highly confidential, while detecting terms related to bid prices, exclusive designs, customer data, and contracts.

Step 2 – Automate post-meeting work. Minutes, decisions, owners, and deadlines are generated immediately after each meeting. Managers can retrieve a decision in minutes instead of searching across multiple applications. A basic control pattern can start like this:

meeting = {
    'sensitivity': 'high',
    'topics': ['bid_price', 'exclusive_design'],
    'participants': ['internal', 'client'],
    'retention_days': 180
}

if meeting['sensitivity'] == 'high':
    require_manager_approval = True
    notify_security_team = True
    revoke_external_access_after = meeting['retention_days']

Step 3 – Block risky actions before execution. HimiTek uses OpenClaw Gatekeeper, its Tool Policy Engine, on 9router v0.4.66 and LiteLLM dual-instance failover. Gatekeeper controls sensitive commands and actions. Elevated tools are locked by default and can run only through a whitelist or explicit user permission.

The Reasoner and Actuator are separated so a prompt injection cannot hijack the entire system. Rate limiting, automatic API key rotation, and hard budget caps, such as 5 USD per month for each virtual key, help prevent runaway loops and unexpected spending. The Agent alerts managers when files become public, are downloaded in bulk, or are shared externally. High-impact actions such as locking, deleting, or revoking access still require human approval.

• Approximately 70% less time spent preparing minutes and tracking action items.
• 30–40 administrative hours saved each month.
• Less manual work for access reviews.
• An estimated VND 360 million saved annually through lower operating effort, fewer errors, and reduced data risks.

4. CTA: Turn meeting data into a controlled business asset

An SME does not need to design a massive project on day one. Start with one focused workflow: automatic minutes, permission control, and alerts for incorrect sharing. Once the results are measured through hours saved, fewer errors, and avoided costs, HimiTek can expand into document governance, customer data protection, and project process control.

Contact HimiTek to assess your current workflow and build a secure meeting-data AI Agent that saves real money, without hiring additional IT staff or continuing to pay for one careless shared link.