A fourteen-person AI agency in Nashville built a document classification system for a mid-size insurance company. The technical work was on schedule. The model performed above the accuracy target. The deployment infrastructure was ready. But the project was delivered six weeks late because the client could not get internal approvals.
The first delay was a two-week wait for the legal team to approve the data processing approach. The second was a three-week wait for the IT security team to approve the deployment architecture. The third was a one-week wait for the business unit head to sign off on the production rollout. Each approval was known to be necessary, but none had been scheduled in advance. Each one was triggered only after the previous step was complete, creating a serial chain of waiting that the agency's delivery team could not influence.
During those six weeks of waiting, three engineers were partially allocated to the project but could not move forward. The agency absorbed $38,000 in underutilized capacity. The client's internal team became frustrated with the delays and began questioning the agency's project management, even though the delays were entirely on the client side.
This scenario plays out at AI agencies constantly. The technical work is done well, but client approval processes create delays that nobody planned for, budgeted for, or managed.
Why Client Approvals Are a Bigger Problem in AI Projects
AI projects interact with more client stakeholders and trigger more internal review processes than traditional software development.
Data governance requires approval. Before an AI system can access, process, or store client data, the client's data governance team typically needs to approve the data processing plan. For regulated industries, this involves legal review and sometimes board-level sign-off.
Model decisions require business validation. Unlike a software feature where the specification defines the expected behavior, model outputs are probabilistic. Business stakeholders need to review model performance, understand the error profile, and approve the system for production use. This approval often requires multiple people who do not sit in the same room.
Deployment requires IT and security sign-off. Deploying an AI system into a client's production environment triggers security reviews, infrastructure approvals, and sometimes change advisory board processes. These reviews have their own timelines that do not align with your project schedule unless you plan for them.
Regulatory compliance requires specialized review. In healthcare, finance, and other regulated industries, AI systems may need compliance review before deployment. These reviews can take weeks and involve external parties.
Executive sponsorship decisions gate major milestones. The executive who approved the project budget often needs to approve major transitions (from discovery to build, from build to deployment, from pilot to production). If that executive is traveling, in budget season, or simply not prioritizing your project, the entire timeline pauses.
Mapping Client Approval Dependencies During Discovery
The time to identify approval requirements is during discovery, not when the team is blocked waiting for them.
During discovery, ask the client:
- Who needs to approve the data processing plan, and what does their approval process look like?
- Who needs to approve the technical architecture and deployment approach?
- Is there a change advisory board or IT governance process that deployments must go through?
- Who needs to approve the model for production use, and what information do they need to make that decision?
- Are there regulatory or compliance reviews required? What are their typical timelines?
- Who is the executive sponsor, and what approvals require their sign-off?
- What is the typical turnaround time for each type of approval?
Document the answers in the project plan. Create a dependency map that shows every approval gate, the approver, the information they need, and the expected turnaround time. Share this map with the client's project sponsor so they understand the approval obligations on their side.
Designing the Approval Workflow
A good approval workflow has three characteristics: it is explicit, it is time-bounded, and it has an escalation path.
Make Approvals Explicit
Every approval should be a named step in the project plan with:
- A clear deliverable that triggers the approval. What does the approver need to review? A document, a demo, a report?
- A named approver. Who specifically needs to approve? Not "the legal team" but "Sarah Chen, Senior Counsel for Data Privacy."
- Acceptance criteria. What constitutes approval? A signed form, an email confirmation, a comment in the project management tool?
- A preparation step. What needs to happen before the approval can be requested? A briefing meeting, a document delivery, a demo session?
Make Approvals Time-Bounded
Every approval should have a defined SLA:
- Routine approvals (status reports, minor deliverables): Two business days
- Technical approvals (architecture, deployment plans): Five business days
- Executive approvals (budget changes, production go-live): Five to ten business days
- Legal and compliance reviews: Ten to twenty business days (plan accordingly)
Include these SLAs in the SOW or project charter so the client has committed to them. Without an agreed-upon timeline, "as soon as possible" means "whenever someone gets to it."
Build Escalation Paths
When an approval is late, your team needs a defined escalation process.
Day 1 past SLA: The project manager sends a polite reminder to the approver with context on the impact of the delay.
Day 3 past SLA: The project manager escalates to the approver's manager or the client's project manager with a note about timeline impact.
Day 5 past SLA: The account manager escalates to the executive sponsor with a formal request to resolve the bottleneck, including the financial and timeline impact of continued delay.
Document this escalation path in the project plan and share it with the client during kickoff. When escalation is expected and pre-agreed, it does not feel adversarial.
Preparing Approval Packages That Accelerate Decisions
One of the most common reasons approvals take too long is that the approver does not have the information they need to make a decision quickly.
Create purpose-built approval packages for each approval type.
Data governance approval package:
- Summary of what data will be accessed, processed, and stored
- Data flow diagram showing where data moves and how it is protected
- Privacy impact assessment addressing relevant regulations
- Retention and deletion plan
- Security controls description
Technical architecture approval package:
- Architecture diagram with component descriptions
- Security assessment covering authentication, encryption, and access control
- Compliance mapping showing how the architecture meets relevant requirements
- Disaster recovery and business continuity provisions
- Cost estimate for ongoing infrastructure
Model production approval package:
- Model evaluation report with performance metrics
- Error analysis showing failure modes and their business impact
- Comparison to baseline or previous system performance
- Monitoring and alerting plan for production
- Rollback procedure if issues are discovered
Each package should include an executive summary on the first page that gives the approver the key information and recommendation in two paragraphs. Many approvers will make their decision based on the summary and only read the details if they have questions. Make it easy for them.
Parallelizing Approvals to Reduce Total Wait Time
If your project has three approvals that each take one week, and you trigger them sequentially, you have three weeks of waiting. If you trigger them in parallel, you have one week.
Identify which approvals can run in parallel. Data governance approval and IT security approval often can proceed simultaneously because they evaluate different aspects of the project. Executive budget approval and technical architecture approval might also be independent.
Create a timeline view that shows parallel approval tracks. Share this with the client so they understand the importance of initiating all approvals as early as possible.
Front-load approval preparation. Start preparing approval packages as early in the project as feasible, even if the information is preliminary. A data governance review that begins with a draft data processing plan during Week 2 is better than one that begins with a final plan during Week 8.
Use conditional approvals when appropriate. "We approve this architecture contingent on the final security assessment confirming the encryption standards" allows work to proceed while the formal process completes. Not every client will accept conditional approvals, but it is worth asking.
Managing Approval Delays When They Happen
Despite your best planning, approval delays will occur. How you manage them determines whether a delay becomes a minor schedule adjustment or a project-derailing crisis.
Track approval status as a first-class project metric. Include approval status in every status report and dashboard update. "Data governance approval: submitted March 3, SLA March 10, status: pending" makes the approval visible and creates a record of the timeline.
Quantify the impact of delays immediately. When an approval is late, calculate and communicate the timeline and cost impact. "Each day this approval is delayed pushes the production go-live back by one day and costs approximately $2,400 in team standby capacity." Concrete numbers motivate action.
Offer to help the approver. Sometimes approvals are delayed because the approver has questions they have not voiced or needs additional information they have not requested. Proactively offer a briefing call: "We are available to walk through the data processing plan with your legal team this week if that would help accelerate the review."
Reallocate team capacity during long waits. If an approval will take three weeks and your team is idle, move them to other projects temporarily. This preserves utilization but requires clear communication about when they will return to the delayed project.
Document everything. When a project is delivered late because of client approval delays, you need a clear record showing that the agency delivered its work on time and the delays were on the client side. This protects you in contract discussions and prevents blame from falling on the delivery team.
Building Approval Awareness Into Your Team
Your delivery team needs to understand that managing approvals is part of managing the project, not a separate administrative concern.
Train PMs to plan for approvals. Approval gates and their timelines should appear in every project plan from day one. A PM who plans the technical work perfectly but ignores the approval timeline will deliver a project that is technically complete but practically late.
Train engineers to flag approval triggers. When an engineer reaches a point where client approval is needed before proceeding, they should notify the PM immediately, not wait until the weekly status call.
Include approval management in PM performance metrics. If a PM consistently has projects delayed by client approvals, either the PM is not managing the approval process effectively or they need help from the account manager. Either way, it should be tracked and addressed.
Contractual Protections for Approval Delays
Your SOW should include provisions that protect the agency when client approval delays extend the project timeline.
Include an approval SLA clause. "Client will respond to approval requests within [X] business days. Delays beyond this SLA will result in a corresponding extension of the project timeline at no additional cost to the client, subject to team availability."
Include a team reallocation clause. "If client approvals are delayed by more than [X] business days, the agency reserves the right to reallocate team members to other engagements. Restarting work after team reallocation may require up to [Y] business days for team reassembly."
Include a project pause clause. "If client-side delays exceed [X] cumulative business days, the agency may pause the project until the client is ready to proceed. A project restart fee of [amount or percentage] may apply."
These clauses are not adversarial. They protect both parties by setting clear expectations about what happens when approvals do not proceed on schedule.
Your Next Step
For your next project kickoff, map every approval gate before the project plan is finalized. Identify the approver, the information they need, the expected timeline, and the escalation path for each one.
Build the approval timeline into the project schedule as explicitly as you build the technical work timeline. Share it with the client during kickoff so everyone understands the approval obligations and their impact on the overall timeline.
Then create your first approval package template for the approval type that most frequently causes delays in your projects. A well-prepared package that gives the approver everything they need on the first try cuts approval time dramatically.
Client approvals are not outside your control. With the right planning, preparation, and process, you can reduce approval delays from weeks to days and keep your projects moving at the pace your delivery team is capable of.