Home

Blog

How Asana Built an AI Engine for GTM with Workato

Customer Stories

Enterprise MCP

Perri Bronson

Sep 30, 2026

Asana is an agentic work management platform with both product-led and sales-led growth motions. In 2024, a small group inside Asana set out to find where AI could help its sales team, surveying reps and surfacing more than 100 potential use cases. Four were chosen as a starting point: competitive intelligence, pre-call planning, post-call planning, and account summaries, all built on Salesforce data with Workato passing information between systems.

What started as a side project became something bigger. Maschal Malek, Product Manager for Enterprise AI at Asana, led the effort to turn that early experimentation into a single AI engine that now powers account research, prospecting, and outreach for the company’s entire sales and customer success organization.

“Our AI engine changes how Asana sells.”

In its first quarter live, the engine drove a 10% increase in bookings, cut account research time from 10 hours a week to 1, and produced 2.3 times the email reply rate of previous outreach.

The Sales Team was Spending Ten Hours Each Week on Research Before a Single Touch

Before the engine existed, Asana’s sales reps were spending roughly 10 hours a week just researching and prioritizing accounts, working across eight or more disconnected tools: 6sense, ZoomInfo, LinkedIn, Chorus, and others, none of which talked to each other.

“What stood out to us most was that our reps were spending about 10 hours per week on prospecting, on account planning, doing research across eight plus different tools, just to figure out their account prioritization,” Malek said. Some reps owned as many as 3,000 accounts, several without any business development rep support at all.

The pain showed up in three places, according to the slides presented alongside the talk:

Account Prioritization: manual scoring, stale enrichment, reactive instead of proactive

Prospecting: tool-hopping to verify contacts and build org charts by hand

Messaging: custom outreach was slow, and reps often turned to unapproved tools like ChatGPT to draft emails because the pre-set sequences from product marketing “aren’t really salesy enough.”

Asana evaluated the AI features being bolted onto existing sales tools already in market. “There were some big name tools that we were using that were like, hey, we could solve all your problems with AI, and we were just like, oh man, the cost is pretty high for these,” Malek said. “None of them really solve our problems… that’s when we landed on we should just use what we already have and put it all together.”

The AI Brain is Powered 90% by Orchestration

The system layers four stages. 6sense, product usage data, ZoomInfo, Salesforce, web research, and Chorus feed a nightly Databricks pipeline that scores every account for new business, expansion, and whitespace fit. That score feeds an AI engine, OpenAI running inside Workato, that generates an account point of view (POV), a prospect POV, email and InMail drafts, and next-best-action recommendations, localized across 19 markets, in roughly 1.5 minutes per account. Workato then writes the output back into Salesforce as a single rich-text field on the account and contact records, so reps see a ranked priority queue, a scored account POV, and a prospect view without leaving their CRM.

“90% of our AI Engine for GTM is powered by orchestration,” Malek said. “A lot of people conflate AI with orchestration… if you don’t have the data sources, AI is not going to give you good responses.”

What Started as a Side Project Became a Lean Revenue-Generating Machine

The team, six people dedicating 15 hours a week each on top of their existing jobs, built what Malek calls Asana’s AI “brain” over four months. The approach: use Workato to connect the tools Asana already owned rather than buying new ones.

The build followed a five-step process: 40+ stakeholder interviews and rep shadowing, a landscape review of the existing tool stack, evaluation of new tools against that landscape, implementation and iteration, and a phased go-live.

Every account POV includes source attribution for each data point, a snapshot of our sales methodology applied to the account (MEDDPICC), competitive battle cards when a competitor like Jira or SmartSheet comes up in a call, and a feedback button that routes directly into an Asana ticket for the build team. Reps can also generate a first-call deck, pre-filled with the target account’s name and relevant use cases, directly from the same interface.

The team rolled out account prioritization, then prospecting, then engagement, one month apart, starting with a 20-person pilot in the corporate sales segment before expanding to the full sales kickoff launch of 260+ reps four months after the project began.

Because everything ran on tools Asana already owned, each generated deliverable, an account POV, a prospect POV, an email draft, cost about 10 cents. “If I went to my CIO at the time and I asked for a shiny new tool, the answer would likely have been no,” Malek said. “At 10 cents per record, nobody was questioning us.”

Making It Agentic with Workato Enterprise MCP

After the core engine was live, the team layered a second capability on top: a custom Model Context Protocol (MCP) server built with Workato, plus a Claude-based plugin with ten skills reps can use to ask account and prospect questions directly in Claude.

Asana built its own MCP rather than adopt an out-of-the-box Salesforce MCP because of the access footprint. “The problem with using a native MCP is it doesn’t necessarily meet your security standards,” Malek said. Asana’s own MCP was scoped to read-only access on the same limited fields the engine already used, with no write-back.

The Claude plugin connects to that MCP and to a Claude project pre-loaded with pitch decks, product FAQs, pricing, and proposal templates, so reps get grounded answers rather than open-ended guesses. Reps can ask for account research, request a competitive battle card, or have Claude draft (but not send) an outreach email, all logged back into Asana as feedback tickets when something needs improvement.

Business Impact – 10% More Bookings, Each Rep Saves 9+ Hours Weekly, 2.3x Reply Rate

Results from the first quarter after the sales kickoff launch:

10% increase in bookings (annual recurring revenue closed) in the first quarter live
10 hours to 1 hour: account research time per rep per week
2.3x higher reply rates compared to prior outreach
86% adoption in the first quarter, since risen to 100%, following adjustments for teams with different account structures (such as enterprise reps managing fewer, larger parent accounts)

Beyond the numbers, Malek pointed to a shift in how the team is perceived internally. “Before we built this, we actually were more of like a serving team… they knew what they wanted, and they would come to us with it,” she said. “We’ve now become thought leaders for our company. We are the ones who are going and finding different areas where we can build for them with AI.”

What’s Next – Sales Plays and a Further Capability Expansion

Since the initial launch, the team has added a customer-success handover workflow (CS Handshake), connected the engine to an e-learning tool for deal support, and built a sales play engine that lets sales ops push targeted messaging and account lists directly to reps’ homepages, for example, surfacing a coordinated outreach play ahead of a conference.

Key Learnings for Building an AI Engine – Start Lean, Build for Scale

Malek closed with four lessons for teams considering a similar build:

Start lean and start with a pilot. The project began as a side effort with a six-person team working 15 hours a week each, not a dedicated build team.
Work with what you’ve got. Asana used tools it already owned rather than buying new AI point solutions, keeping the marginal cost near zero.
Be open to feedback and iterate. Recipes changed multiple times a day during rollout based on direct rep feedback.
Build for scale from the start. An earlier, narrower version of the solution had to be rebuilt repeatedly as new teams asked for it; building the shared “brain” architecture up front avoided that cycle.

See Workato in Action

Asana’s team built a 10% bookings lift, 90% less time on account research, and a 2.3x reply rate improvement using tools they already owned, orchestrated through Workato.

Learn how you can build your own AI engine. Schedule a demo →