The agent is only as good as its data
A customer mentions a new team on a sales call. The recording sits in Gong, the account lives in Salesforce and the next step gets discussed in Slack.
An agent needs those pieces together before it can recommend an expansion opportunity or update the account. Kiwi builds and maintains that shared record for go-to-market teams. [1]
Kiwi calls itself your GTM agent’s favorite data engineer. It brings the data, agent instructions and approvals into one system, then uses each run to improve the next. [1][2]
One record of the account
Kiwi syncs data from tools such as Salesforce, HubSpot, Gong, Slack, Gmail and Granola into a versioned GTM data lake. [1][2]
The team explains a job in plain language. Kiwi builds the agents and steps behind it, using that shared record to answer questions or prepare actions. [1]
Every data-backed answer can point to the SQL query and data snapshot used to produce it. Teams can trace a number back to its source and replay a run against the same record. [1][2]
Kiwi also checks whether required data is fresh. When a run needs an update, it can fetch fresh information instead of continuing with an old account picture. [1]
Turn a call into a reviewed next step
One Kiwi demonstration starts with expansion signals in Gong calls: a customer mentions hiring, another team or new budget. [1]
Kiwi finds the signal and brings the supporting quote to a person for review. Once confirmed, the agent can prepare the Salesforce update and the message that tells the account team what changed. [1]
Each action waits for approval. Kiwi does not send an email, create an event or write to connected tools until a person approves that specific step. [2]
The same approach supports renewal preparation, account timelines, pipeline reviews and turning meeting notes into Linear tickets. The source material stays attached to the work the agent prepares. [1]
The review becomes part of the system
An approval, edit or rejection tells Kiwi how the team wants that job done. Kiwi records those decisions and uses them to test and improve the agents. [1][2]
Historical runs become evaluations for future changes. Teams can check a revised agent against decisions they have already made, rather than relying on a fresh prompt alone. [1]
Kiwi also maintains agent skills as the business changes. Data connections, reusable instructions and the record of past decisions stay with the recurring job. [1][2]
Built by people who have maintained the data
Melanie Chen led post-sales at Chalk after joining as a founding forward-deployed engineer. Earlier, she built data infrastructure at Plaid. [2]
As engineering teams adopted agents, she kept encountering the data and maintenance work that held other teams back. In one week, she spent 20 hours fixing GTM agent work. [3]
Bill Qin brings a similar infrastructure background, including building Chalk SQL and quantitative research at Jane Street. Together, they are building Kiwi for the teams that need agents to keep working after the first demonstration. [2]
GTM teams can book a demo to explore their own tools and recurring jobs. [1][2]
Where Kiwi ranks
Kiwi isn't in a numbered ranking yet. Products get a position once enough companies compete on the same task.
Current rankings. Positions come from public evidence, and no company can pay for a place. How we rank
The bigger picture
A sales team needs an account update it can act on, with the source behind it and a person controlling what goes out. [1][2]
Kiwi is building the data engineering and maintenance underneath that experience, so the agent can become part of the team’s everyday work. [1]
Written by SOTA2 from Kiwi’s public materials. Customer experiences and product plans are reported by Kiwi.