Scaling Meta & Google Ads to ₹12Cr+ spend with AI optimisation

Boss Digital managed performance marketing campaigns for 40+ clients across e-commerce, real estate, education, and healthcare verticals. As client budgets grew, the limitations of manual campaign management became acute: bid adjustments that needed to happen in real time were being made daily or weekly, creative fatigue was reducing ROAS as the same creatives ran for weeks without rotation, and the account management team was stretched across too many clients to give any single account the attention that maximised performance.

The agency was losing client renewals to larger competitors who could demonstrate AI-driven optimisation capabilities. Several enterprise clients had explicitly asked about automated bidding beyond Meta's built-in Advantage+ features and about creative personalisation at scale. Boss Digital's technical stack — built on standard agency tools — had no capability to deliver on these requirements, putting their client retention and new business pipeline at risk.

Creative production was the single biggest bottleneck. Each client required regular fresh creative assets — static images, carousel ads, video ads, copy variants — to prevent performance decline from ad fatigue. The in-house design team could not produce enough volume to support meaningful creative testing across 40 clients simultaneously. The agency was effectively leaving optimisation gains on the table because creative velocity was too slow to test at the frequency that modern performance marketing requires.

We built a custom AI campaign optimisation layer that sat on top of Meta and Google Ads APIs, operating at a granularity and frequency that the platforms' own automation tools could not match. Custom bidding algorithms evaluated conversion probability signals — time of day, device, audience segment, creative fatigue score, and historical conversion patterns — and adjusted bids every four hours rather than the platform default of daily. For Google Performance Max campaigns, we built automated asset group rotation based on performance signals.

The creative automation pipeline used a combination of generative AI (GPT-4 for copy, Stable Diffusion fine-tuned on each client's brand assets for visuals) and a structured templating system to produce hundreds of creative variants per client per week. A quality review workflow ensured brand compliance before any AI-generated creative reached the platform. Performance data from each creative was fed back into the generation model, progressively refining what it produced based on what was actually converting.

A unified reporting dashboard was built for Boss Digital's account management team, consolidating cross-platform performance data for all 40 clients into a single view. Anomaly detection algorithms flagged accounts where performance had moved outside expected ranges — whether up or down — so account managers could identify and amplify wins as quickly as they could catch and fix problems.

The AI bidding layer was deployed across 28 of the agency's highest-spend client accounts, managing a combined monthly budget of ₹12 crore+. The system executes over 50,000 individual bid adjustments per day across all accounts — a volume of optimisation activity that would require a team of 15+ dedicated specialists to replicate manually. Average CPA across the managed portfolio fell 35% in the first three months as the bidding model accumulated sufficient data to make high-confidence adjustments.

The creative automation pipeline now produces 800-1,200 unique creative variants per month across all client accounts. An A/B testing framework automatically allocates budget toward winning creatives and pauses underperformers, with statistical significance calculated at the ad set level rather than the campaign level for faster, more reliable conclusions. Creative fatigue — measured by declining CTR and CPM inflation — has been reduced from an average 14-day creative cycle to a 4-day average refresh cadence.

Boss Digital's account management team now handles 40% more client accounts per head than before the engagement, with no reduction in client satisfaction scores. The AI capabilities have become a significant competitive differentiator in their new business pitches — in the 12 months since deployment, the agency has closed six new enterprise clients who specifically cited the AI optimisation offering as a key decision factor, adding ₹3.2 crore in annual recurring revenue.