The Business Challenge
The business situation the organisation faced before the project began.
Creating high-converting influencer campaigns on Influns previously took 1 to 2 weeks per brand. The manual workflow required extensive research across social platforms, analyzing recent brand collaborations, evaluating digital presence, conducting competitor benchmarking, and running creative brainstorming sessions. This heavy reliance on human labor created severe operational bottlenecks, limiting platform scalability, increasing turnaround times, and restricting the volume of campaigns a brand could run.
Why AI Was the Right Solution
Why AI was the right approach — and what alternatives were considered.
Traditional rules engine or template-based software could not handle the unstructured, highly dynamic nature of social media content, brand aesthetic evaluations, or multi-channel competitor positioning. Deterministic approaches lack the contextual reasoning required to synthesize historical post performance into actionable creative strategies. Generative AI and multi-agent orchestration provided the exact blend of reasoning, semantic understanding, and creative synthesis needed to emulate human strategic thinking at scale.
How the Solution Was Delivered
Discovery, design, development, testing and rollout — the journey, not the tooling.
- 1Discovery & brand workflow mapping
- 2Gold-set strategy and brief calibration
- 3Multi-agent pipeline design on Vertex AI & Google Gemini
- 4Asynchronous queue implementation with Google Pub/Sub, Tasks & Go
- 5Shadow-mode validation against historic human briefs
- 6Assisted-mode launch for campaign strategists
- 7Continuous feedback integration and output scoring
Key Technical & Architecture Decisions
Architecture, model selection, workflow and trade-offs.
• Multi-Agent Specialization over Single-Prompt Execution: Decomposed the workflow into five specialized agents (Brand Analysis, Footprint Discovery, Competitor Benchmarking, Strategic Synthesis, Campaign Ideation). This modularity isolated failures and allowed individual model fine-tuning and prompt calibration per task.
• Vertex AI & Gemini: Selected Google Gemini on Vertex AI for superior contextual understanding across long unstructured data inputs and multimodal brand asset analysis.
• Asynchronous Microservices Architecture: Standardized background orchestration using Go and Google Pub/Sub / Tasks. This decoupled long-running research tasks from user sessions, enabling parallel agent processing without UI blocking.
Challenges & How They Were Solved
Obstacles hit along the way and how they were overcome.
• Context Drift Across Agents: Early prototypes lost context when passing strategy outputs to the brief generation agent. Solved by defining strict JSON schemas for inter-agent communication, ensuring upstream insights directly informed downstream deliverables.
• Inconsistent Competitor Signals: Web and social data contained noise and irregular formats. We implemented standard filtering and structured extraction pipelines prior to feeding data into the Benchmarking Agent.
User Adoption & Change Management
How users responded — training, change management and feedback loops.
• Shadow-Mode Testing: Deployed the multi-agent pipeline in shadow mode alongside human strategists for three weeks to compare AI-generated briefs against historic benchmarks without impacting live accounts.
• Human-in-the-Loop Workflow: Designed the UI to present strategists with three actionable strategies first. Humans retain complete control to select, reset, or tweak the strategic direction before triggering the final brief creation.
Business Outcomes & Impact
The measurable outcomes from the underlying AI Deployment — with the story behind them.
Reduced end-to-end campaign creation and brief generation time from 1–2 weeks (80+ hours) down to 15 minutes per campaign, achieving a >98% reduction in cycle time. Completely automated manual research across brand history, competitor intelligence, and market trends, eliminating operational bottlenecks and allowing the platform to scale campaign output without increasing strategy headcount.
Business Outcome Categories
The story behind the numbers
Reducing campaign creation time from 80+ hours down to 15 minutes unlocked immediate operational scale without increasing strategy headcount. Marketers and agency clients transitioned from spending days on manual research to focusing on high-level creative direction and influencer relationships. By automating brand history analysis and competitor benchmarking, campaign briefs became substantially more data-backed and consistent, driving higher influencer acceptance rates and accelerating client onboarding.
Lessons Learned
What surprised the team, what worked well, and what would be done differently.
- Breaking complex strategy into specialized agents improves quality dramatically over single-prompt workflows.
- Structuring inter-agent data schemas early prevents context drift.
- Human-in-the-loop validation builds trust faster during initial deployment.
Future Opportunities
Where this solution could go next.
• Automated Influencer Matching: Integrating performance analytics directly into candidate influencer selection models.
• Closed-Loop Performance Learning: Feeding real-time campaign engagement results back into the Strategic Synthesis Agent to continuously refine future strategy generation models.
Professional Reflections
Guidance for another builder tackling a similar problem.
• Focus on Workflow Deconstruction: Breaking down complex human tasks into distinct, micro-scoped agent responsibilities yields far better quality than attempting a single, comprehensive prompt.
• Invest Early in Schema Validation: Enforcing structured JSON contracts between agents prevents cascading errors across multi-step agentic pipelines.