How Push Automated 95% of Its Sales Process With AI

Overview

For Push, AI wasn’t an experiment. It was a structural decision.

Push is a UK-based marketing agency working across campaign delivery, training, and consultancy. Like most agencies, their work depends on speed, consistency, and execution under pressure. The difference is how early they recognized that AI wasn’t just going to make that work faster. It was going to change how it needed to be done.

When ChatGPT surfaced publicly in November 2022, Push didn't treat it as something to test quietly or explore on the side. It didn't arrive through a roadmap or a pilot program. Someone shared it in the agency's internal WhatsApp group with a simple message encouraging everyone to try it.

Within a week, leadership aligned. As a Microsoft Partner with beta access, Push had an early advantage. When the rest of the industry began paying attention, Push was already moving.

They didn't wait to see how things would develop.

Senior team leads were pulled into a boardroom and handed a clear mandate: the era of "business as usual" is over. Assume AI is the baseline for your workflow, and redesign your entire process from the ground up.

As Jai Khan, who leads sales at Push, recalls, that moment set the tone for everything that followed.

“Someone sent ChatGPT into the WhatsApp group and said, ‘This thing looks amazing and everyone should try it out.’ And the following week, our CEOs brought everyone, all the senior team leads, into a boardroom meeting to say, ‘All of you need to redevelop your entire processes to use as much AI as possible.’”

From that point on, AI wasn’t a side project at Push. It was something the agency would build around.

But making that directive real required more than enthusiasm. It required structure.

From Early AI Use to Real Structure

At first, the motivation was practical.

Teams wanted to move faster. Sales wanted better research before calls. Marketing wanted quicker analysis. Everyone wanted less time spent copying, pasting, and stitching information together across too many tools.

The focus was internal.

How can this help people do their jobs better? That pattern showed up consistently across the agency.

“Initially it was about saving time and money,” Jai said. “At the beginning it was just like, what can be done quicker and faster for ourselves? I don’t think we thought about the implications of how this could actually help our clients.”

AI helped, but only in moments. Prompts lived in isolation. Outputs varied. Work still moved one step at a time.

If AI was going to matter long term, it needed to stop being something people used occasionally and start becoming something work flowed through.

“Once people started using it more, the thinking changed,” Jai said. “The more we used AI, the more creative and ambitious the use cases became. It stopped being about one task at a time and started becoming about entire processes.”

That realization is what led Push to Cassidy.

Before adopting Cassidy, the team had already built an internal AI platform to use language models securely. It worked, but it was rigid. Agents were difficult to extend. Not everyone could build or modify them. And most importantly, the system couldn’t support work end to end.

According to Paul Kearney, Managing Director at Push, Cassidy changed that.

“Organizations need a comprehensive, yet simple AI operating system” Paul said. “Cassidy is the perfect operating system.”

What stood out immediately wasn’t just the Agents themselves. It was the ability to connect them into real workflows.

“The first, most immediate difference with Cassidy was actually the Workflows,” Jai said. “Being able to create different Agents and chain them all together was a game changer.”

That shift marked the moment AI stopped being a speed boost and started becoming structural.

Sales was where that change showed up first.

Rebuilding the Sales Engine Inside Cassidy

Once Push saw what was possible with connected Workflows, the focus shifted quickly from individual efficiency to team enablement.

The goal wasn’t to give one seller an edge. It was to standardize how the entire sales team prepared, pitched, and followed up, without increasing headcount or complexity.

Push rebuilt the sales process inside Cassidy step by step, designing Workflows that would trigger automatically and support anyone touching an opportunity.

Cassidy was integrated directly with HubSpot, so the moment a lead is marked as an opportunity, Workflows activate for the whole team. From there, the system begins preparing everything sales needs before a conversation even starts.

Before the first call happens, Cassidy handles the heavy lifting:

• Researching the company and its market
• Analyzing competitors and digital strategy
• Reviewing website content and messaging
• Scraping sites to generate tone-of-voice documentation
• Building LinkedIn-based personality profiles for key stakeholders
• Identifying personas, pain points, and purchase considerations
• Preparing a full sales pitch in advance

By the time a seller joins the call, the groundwork is already done. Preparation is no longer dependent on individual habits or experience. Everyone starts from the same baseline.

“Before that first call, we want all the research done,” Jai explained. “Everything the team needs should already be there. That way, anyone can walk into a conversation fully prepared without spending hours pulling things together.”

After the call, Cassidy continues where most systems stop.

Call transcripts are processed automatically, and structured briefs are generated for delivery teams. Notes are standardized. The next steps are clear. Nothing gets lost between sales and execution.

Push also built a sales coaching Assistant that reviews call transcripts and provides consistent feedback across the team. Instead of coaching depending on availability or memory, sellers receive immediate, objective insights they can act on while conversations are still fresh.

“Being able to get feedback right after a meeting and understand how it went is something we never really had before at scale,” Jai said.

The result is near-total automation of the sales workflow, not for one person, but for the organization.

“When we talk about how we’re using Cassidy, the reality is the sales process is about 95% automated now. Everything happens the minute an opportunity comes into the business.”

That level of consistency changed how Push showed up in conversations. It also changed outcomes.

This year, Push closed roughly 70% of deals. The team credits part of that success to being able to demonstrate real, working AI workflows in the sales process, not just talk about AI in theory.

Removing Weeks of Work From Core Agency Processes

Once sales was working differently, the next question became obvious.
If Workflows could replace weeks of effort there, where else was time still being lost?

RFPs surfaced immediately.

They had long been some of the most resource-intensive projects Push handled. Large documents. Multiple contributors. Long review cycles. Entire weeks spent coordinating input and assembling responses under pressure.

Cassidy allowed Push to rebuild that process end to end.

The team trained an RFP Assistant on historical responses and internal documentation. When a new RFP arrives, the system now generates a complete first draft in a single day, structured and ready for review.

As Jai described it:

“Normally it would be weeks across the team, with four or five people spending three or four days each. We built an RFP Assistant trained on all of our old responses and turned the whole thing around in one day.”

Cassidy then assigns clear review ownership across the team, eliminating version chaos and back-and-forth. They won the RFP.

That win made something clear internally. The same workflow approach could be applied to other high-effort, high-friction areas of the business.

Across the agency, teams began rebuilding processes that had previously relied on manual effort:

• Ongoing competitor analysis, automated to run weekly and monitor website changes, messaging shifts, and social strategy updates
• Marketing research Workflows that consolidated industry analysis into repeatable, structured outputs
• Performance analysis for Google Ads, where a specialist chained more than 30 Agents together to produce 100+ page deep-dive account reports from raw platform data

What once required hours of manual checking and interpretation now happened continuously in the background.

That internal proof point mattered. AI wasn’t just helping teams move faster. When structured properly, it could support highly technical, end-to-end workflows that had previously relied entirely on human effort.

Making AI Part of Everyday Work at Push

Once AI workflows were running across sales, marketing, and performance teams, the challenge at Push shifted.

It was no longer about whether AI could help. It was about whether it would become something people actually used every day.

Getting a handful of power users onboard is easy. Getting an entire agency to rely on AI as part of daily work is much harder.

What made the difference at Push was how intentionally adoption was handled.

AI wasn’t mandated from the top down. Teams weren’t handed rigid tools and told to comply. Instead, people were encouraged to build their own Agents and Workflows in ways that matched how they already worked.

Cassidy’s sandbox environment played a key role. It gave team members space to experiment without risking live systems, while secure folders, access controls, and siloed knowledge bases ensured sensitive information stayed protected.

“Having something where you can be the master of your own destiny and create your own Agents really easily encouraged adoption,” Jai said. “People that were nervous about AI felt more comfortable because they could test things in a sandbox without affecting the wider company.”

As more people started building and sharing Workflows, AI use spread organically. It stopped being something a few individuals were good at and became something teams relied on collectively.

Support played an equally important role in making that stick.

Push evaluated multiple platforms before choosing Cassidy. Integration depth and usability mattered, but what stood out over time was how present Cassidy’s team remained once real usage began.

“I kind of expected after onboarding we’d be left to it,” Jai said. “But Cassidy’s team has been on calls almost every week helping us get Workflows working and corrected.”

That ongoing support mattered because questions didn’t come from beginners. They came from people actively building, extending Workflows, and pushing the system into more complex use cases.

Instead of slowing down when things got harder, teams had help refining and improving what they had already created.

The result was sustained, agency-wide adoption.

AI didn’t live with a single champion at Push. It became part of how work actually got done day to day. People didn’t have to remember to use it. It was already embedded in their Workflows.

AI Became the Agency’s Operating System

Push doesn’t see Cassidy as a tool it could swap out or phase down.

“There’s no way we could ever go back to not using a system like this,” Jai said. “It’s fundamentally changed the way everyone in the organization works.”

AI isn’t an experiment at Push. It’s infrastructure.

Their story shows what happens when AI adoption is intentional, structured, and embedded into real workflows. Cassidy helped Push move beyond experimentation into execution, turning ideas into repeatable systems across sales, marketing, and operations.

If your team is serious about scaling AI in a way that actually sticks, this is what it looks like in practice.

Book a demo to see how Cassidy can support your Workflows.