I recently sat down with Ben Robbins, talent partner at Tandem Health, to see what AI agents look like when the open tabs are LinkedIn, Ashby and Slack rather than a sanitised demo.
Tandem is one of Stockholm's leading AI scale-ups, helped in no small part by the phenomenal work Ben and its talent team are doing. This creates a problem for me: I compete with them every time Strawberry hires. The silver lining is that they run much of this work in Strawberry Browser, which at least evens things out. Slightly.
This is the important bit: Strawberry is not another chat window sitting next to the work. It is a browser with AI Companions built in, so Ben can move through logged-in recruiting tools, keep Tandem's hiring context with the workflow, and turn work that earns trust into a shared Skill or scheduled Routine.
Inside Ben's talent workflows
The easiest way to start is with one annoying task in the tab you already have open. Your Companion can use the role, ATS tab and hiring context already in front of you.
/getting-started-with-recruiting-in-strawberry Find people a keyword search misses
Ben uses Strawberry predominantly for sourcing. A normal LinkedIn or Boolean search asks him to turn recruiting judgement into exact words first. His Companion can work from the role and Tandem's shared sourcing rules, move through LinkedIn in the visible browser, and follow signals that matter even when a profile does not contain the obvious keyword.
Rather than relying on brittle keyword hits, Strawberry evaluates the live LinkedIn profile against the role and Tandem's shared sourcing rules. The useful output is not a longer list. It is a candidate set with the reason each person belongs, the profile evidence the recruiter can inspect, and the uncertainty that still needs a human view.
Before Strawberry, Ben would spend 60 to 90 minutes on Monday mornings in LinkedIn, connecting with people who were broadly relevant to his work or specifically relevant to an open role. The next part of his setup decides where that attention should go before he opens the search.
Let Ashby set the Monday priority
Every Monday morning, Ben has a Strawberry Routine look at the surface-level pipeline data in Ashby and tell him where the pipeline is thin. The Routine does more than count names. A role with two candidates at final stage can be healthier than one with 20 candidates still at application review, so it weighs where the candidates are before suggesting which role deserves sourcing attention.
This is where the browser matters. The recommendation is made from the live recruiting context rather than a pipeline summary pasted into a separate chat. Ben can inspect what Strawberry found in Ashby, then continue into the LinkedIn work the recommendation calls for. Because it runs on Monday morning, he does not have to remember to rebuild the same review each week.
Put the scorecard inside the search
Ben started by learning how to prompt Strawberry and customise the work for himself. He then turned Tandem's sourcing non-negotiables into shared rulebooks. Two of the rules he gave in our conversation were to exclude anyone currently working at Tandem and anyone who had interviewed there in the previous six months.
Those rules now travel with the sourcing workflow. Ben says each run tells him which prospects it picked up and which it discarded because of the rules, so the recruiter can review both the shortlist and the judgement behind it. The context does not have to be explained again at the start of every search.
The same playbooks are shared across Tandem's talent team. Ben has also created an onboarding guide for new talent hires, covering how the team works and how they use Strawberry. A new joiner starts with the team's context and methods rather than Ben's private prompt history.
Protect the candidate experience in Slack
Ben also set up a candidate-experience SLA. If someone has been sitting too long at a stage in Ashby, a Strawberry Routine pings him in Slack and tells him that the candidate needs attention. It is a small piece of admin with a much bigger consequence than its size suggests.
Tandem is growing, and Ben sees every candidate interaction as a chance to create a promoter or a detractor whether or not that person gets the job. The Routine does not make a hiring decision. It checks the pipeline against the team's timing rule and puts the exception where Ben will see it, while there is still time for a human to follow up.
Start with the task you want to stop doing
Ben didn't start with an elaborate master plan. He started using Strawberry himself, corrected the prompts and rules, then expanded the workflows to his team. After five or six months, he had six or seven Routines running. The sequence matters more than the number: prove the work, keep the judgement that improves it, share it and only then put the stable repetition on a schedule.
Give Strawberry the repetitive talent task you most want to stop doing. It might be the Monday pipeline review, a sourcing search LinkedIn keeps missing, or the check for candidates who have waited too long. Bring the relevant tabs and team context; your Companion can do the checking in the same browser where the work lives and leave you with something you can inspect, correct and reuse.
Unless you recruit in Stockholm. I already have enough competition.