A 44-person AI agency in Minneapolis ran their first comprehensive tech stack audit and found 63 active software subscriptions totaling $31,000 per month. The number shocked the leadership team — they had estimated around 35 tools and $18,000 in monthly spend. The gap existed because tools were purchased across multiple departments without central oversight, some subscriptions had been started by employees who had since left, and several tools had been adopted for specific projects and never cancelled after the project ended. The audit identified $9,800 per month in immediate savings — tools with zero active users, duplicate tools serving the same function, and premium tiers on tools that only needed basic plans. That was $117,600 per year returned to the bottom line with no impact on productivity. After optimizing the remaining tools and negotiating renewals, they saved an additional $4,200 per month. Total annual savings: $168,000 — enough to fund a senior engineer's salary.
Tech stack audits are one of the highest-ROI operational activities an agency can undertake. They are also one of the most neglected, because nobody owns the tech stack holistically. Engineering owns development tools. PM owns project management tools. Marketing owns marketing tools. Finance owns accounting tools. Nobody looks at the whole picture. That fragmented ownership is exactly how tool sprawl happens.
Why Tech Stacks Become Bloated
The "Try It and Forget It" Pattern
Someone on the team signs up for a tool to solve a specific problem. They use it for a few weeks, decide it is not quite right, and switch to something else. But they never cancel the original tool — they just stop logging in. The subscription charges continue silently.
The "Every Team Picks Their Own" Pattern
Without a centralized tool policy, different teams adopt different tools for the same function. One project team uses Notion, another uses Confluence, a third uses Google Docs. Each tool has its own subscription, and the agency's knowledge is fragmented across all three.
The "Upgrade and Never Downgrade" Pattern
During a busy period, someone upgrades to a premium tier to access a specific feature. The busy period ends, the feature is no longer needed, but nobody downgrades. The premium charges continue for months or years.
The "Free Trial to Paid" Pattern
Free trials that automatically convert to paid subscriptions are designed to capture exactly this revenue. Someone starts a trial, uses the tool briefly, and either forgets about the trial or does not realize it converted to paid. The tool charges against a corporate card that nobody reviews line-by-line.
The "Legacy Tool" Pattern
The agency adopted a tool three years ago. Since then, better options have emerged, or other tools in the stack now include that functionality. But migration seems like a hassle, so the old tool stays. Meanwhile, new team members are confused about which tool to use for what purpose.
The Tech Stack Audit Process
Phase 1: Inventory (Week 1)
Build a complete inventory of every software subscription, cloud service, and technology tool your agency uses.
Data sources for inventory:
- Credit card and bank statements: Review the last 12 months of charges for recurring subscriptions. This catches tools purchased on personal or department cards
- Email records: Search for subscription confirmation emails, renewal notices, and billing receipts
- Expense reports: Review expense reports for reimbursed software purchases
- IT records: Check your identity provider (Google Workspace, Okta, etc.) for SSO-connected applications
- Cloud provider dashboards: Review AWS, GCP, and Azure billing for all active services
- Team survey: Ask each team member to list the tools they use daily, weekly, and occasionally
For each tool, capture:
| Field | Description | |-------|-------------| | Tool name | Name of the software or service | | Vendor | Company providing the tool | | Category | Function it serves (project management, communication, development, etc.) | | Monthly cost | Current monthly charge | | Annual cost | Total annual cost including any annual subscriptions | | Payment method | Which credit card or billing account | | Plan/tier | Which plan you are on (free, basic, pro, enterprise) | | Contract end date | When the current contract or subscription term ends | | Auto-renewal | Yes/No and notice period for cancellation | | Licensed users | Number of paid seats/licenses | | Active users (30 days) | Number of people who actually logged in during the last 30 days | | Active users (90 days) | Number of people who logged in during the last 90 days | | Primary team | Which team uses this tool most | | Owner | Who manages the account and is responsible for the tool | | Integration points | Which other tools does it connect with |
Phase 2: Analysis (Week 2)
With the inventory complete, analyze it across several dimensions.
Usage Analysis:
Flag any tool where:
- Zero users in 90 days (immediate cancellation candidate)
- Active users are less than 50% of licensed users (right-size the plan)
- Usage has declined over the last 6 months (assess whether the tool is still needed)
Overlap Analysis:
Map tools by function and identify overlaps:
- How many tools serve project management?
- How many tools serve communication?
- How many tools serve documentation/knowledge management?
- How many tools serve design?
- How many tools serve time tracking?
- How many tools serve file storage?
Any function served by more than one tool is a consolidation opportunity. The goal is one primary tool per function, with rare exceptions for tools that serve genuinely different use cases.
Cost-Value Analysis:
For each tool, assess the value it provides relative to its cost:
- High value, low cost: Keep. These are your workhorses — tools that are heavily used and reasonably priced
- High value, high cost: Optimize. Negotiate pricing, right-size the plan, or evaluate alternatives that provide similar value at lower cost
- Low value, low cost: Evaluate. These might not be worth the cognitive overhead of maintaining them even if the cost is minimal
- Low value, high cost: Eliminate or replace. These are your biggest waste targets
Integration Analysis:
Map how tools connect to each other. Tools that are deeply integrated with your workflow are harder to replace but also more valuable. Tools that operate in isolation are easier to eliminate but also suggest they are not essential.
Security and Compliance Analysis:
Check each tool against your security requirements:
- Does it have appropriate security certifications (SOC 2, GDPR compliance)?
- Does it store sensitive client data? If so, does it meet your data handling requirements?
- Is it approved for use with client data, or has someone been using it without approval?
- Are access controls properly configured (SSO, 2FA)?
Phase 3: Action Plan (Week 3)
Based on the analysis, create an action plan with four categories:
Immediate Cancellations: Tools with zero or near-zero usage. Cancel these this week. If anyone complains, you can always resubscribe.
Right-Sizing: Tools where you are paying for more licenses or a higher tier than needed. Downgrade at the next billing cycle.
Consolidation Projects: Functions served by multiple tools that should be consolidated to one. For each consolidation:
- Choose the winning tool based on feature coverage, user preference, and cost
- Plan the migration (data transfer, workflow changes, training)
- Set a cutoff date for the retired tool
- Communicate the change to affected teams
Negotiation Targets: Tools that are valuable but expensive, or contracts that are up for renewal. Prepare for negotiation:
- Research competitor pricing
- Identify your leverage (contract expiration, usage data, competitor quotes)
- Set a target price reduction
Phase 4: Execution (Weeks 4-8)
Execute the action plan in order of impact:
Week 4: Cancel unused tools and process downgrades Weeks 5-6: Begin consolidation projects (choose tools, plan migration) Weeks 6-7: Execute migrations and train teams on consolidated tools Week 8: Complete negotiations on remaining tools, sunset retired tools
The Ideal AI Agency Tech Stack
While every agency's needs differ, here is a reference architecture for a 20-50 person AI agency:
Communication and Collaboration
- Slack or Microsoft Teams — real-time messaging and channels (pick one)
- Google Workspace or Microsoft 365 — email, calendar, documents, spreadsheets (pick one)
- Zoom or Google Meet — video conferencing (often included in your workspace suite)
- Loom — async video communication
Project Management and Delivery
- Jira, Linear, or Asana — project and task management (pick one)
- Confluence, Notion, or GitBook — documentation and knowledge management (pick one)
- Figma — design collaboration (if your agency does any UI/UX work)
Development and Engineering
- GitHub or GitLab — code repository and CI/CD (pick one)
- VS Code or JetBrains — development environment (individual choice is fine here)
- Docker — containerization
- Terraform or Pulumi — infrastructure as code
ML and Data
- MLflow or Weights & Biases — experiment tracking and model registry (pick one)
- Jupyter/JupyterHub — notebooks for exploration and analysis
- dbt — data transformation (if your agency does significant data engineering)
- Airflow or Dagster — workflow orchestration (pick one if needed)
Cloud Infrastructure
- AWS, GCP, or Azure — primary cloud provider (pick one as primary, with secondary for client requirements)
- Kubernetes — container orchestration (if at sufficient scale)
Business Operations
- QuickBooks Online or Xero — accounting
- Harvest or Toggl — time tracking
- Brex or Ramp — corporate cards and expense management
- Gusto or Rippling — payroll and HR (pick one)
- HubSpot or Salesforce — CRM (pick one)
Security and Compliance
- 1Password or Bitwarden — password management
- Vanta or Drata — compliance automation (if pursuing SOC 2)
- Cloudflare — web security and performance
Total tool count: approximately 20-25 tools. If your count exceeds 40, you almost certainly have overlap and waste.
Building Ongoing Tech Stack Governance
The Tool Approval Process
Implement a lightweight approval process for new tool purchases:
Under $50/month: Team lead can approve. Notify operations for tracking. $50-500/month: Operations approval required. Must demonstrate that no existing tool solves the problem. Over $500/month: Leadership approval required. Requires a brief evaluation document comparing alternatives.
Quarterly Usage Review
Every quarter, pull usage data for your top 20 tools by cost. Review:
- Are usage levels consistent with the previous quarter?
- Are there any significant drops in usage?
- Are there new tools that have appeared without going through the approval process?
Annual Tech Stack Audit
Run the full audit process (inventory, analysis, action plan, execution) annually. Even with good governance, tools accumulate. The annual audit is your reset.
Vendor Relationship Management
For your most important vendors (top 10 by spend), maintain active vendor relationships:
- Know your account manager and have their direct contact
- Understand your contract terms, including renewal dates and cancellation windows
- Track vendor roadmaps to ensure the tool will continue meeting your needs
- Negotiate from a position of knowledge, not desperation, at renewal time
Measuring Tech Stack Health
- Total monthly tech spend — track over time, both absolute and as a percentage of revenue
- Tools per person — total unique tools / total team members. Target: 8-12 tools per person
- License utilization — paid licenses / active users. Target: over 80%
- Overlap count — number of functions served by more than one tool. Target: zero (or minimal with documented justification)
- Tool satisfaction score — quarterly survey of team satisfaction with the toolset. Low satisfaction indicates adoption problems or poor tool choices
Your Next Step
Start the inventory this week. Pull your credit card statements for the last three months and list every recurring software charge. You do not need to complete the full audit in one sitting — just building the inventory will reveal tools you forgot you were paying for. Then pick the five most expensive tools and check their usage data. If any of those five has significantly fewer active users than paid licenses, that is your first optimization target. Right-size it this month. The savings from those five tools alone will likely be meaningful, and the exercise will motivate you to complete the full audit.