Cold Outbound SDR Agent
Builds a fresh, verified batch every week
Rhythm Mondays 6:45am lead pull, reply checks twice a weekday
GoDataFeed sales · Team overview
Six AI agents we built to run the repetitive half of GoDataFeed's sales work: finding companies with a reason to talk, writing tailored outreach, watching every reply and keeping HubSpot honest. Benjamin approves anything that reaches a prospect.
Live numbers, latest report —
How the team works together
Every agent has its own district that grows as the agent gets results, measured against its own usual week. Vans carry real work along the streets into HubSpot and on to customers. Tap a district to fly in, slide back in time to see how it grew, or tap the i to see how the city grows.
Meet the team
Each agent owns one part of the sales motion. Open any card for its full tab.
Cold Outbound SDR Agent
Rhythm Mondays 6:45am lead pull, reply checks twice a weekday
Warm Signal SDR Agent
Rhythm Weekdays 7:45am intent pull, Tue displacement, Mon Second Look, every email verified first
Sales Nav Agent
Rhythm Agencies Mondays, brands Wednesdays, ready about 8:10am
Instantly Agent
Rhythm Syncs every weekday at 3:32pm (live campaigns in full, new replies on finished ones), audit monthly
HubSpot Lead Agent
Rhythm Weekdays 8:10am sweep, 3:25pm check-in that also posts its numbers here
LinkedIn Connect Agent
Rhythm Reply checks 8:20am and 3:25pm, one combined connect batch Fridays 12:47pm with a backstop
Week by week
New people each agent contacted personally, by the week it happened. Instantly campaign volume is counted separately above.
Working side by side
Rule 1 · Ownership
The LinkedIn Connect agent does all LinkedIn work. When someone another agent found accepts or replies, it logs it on both tabs and hands it back to that agent, which also checks its own tab, so no reply sits unanswered.
Rule 2 · Screen first
Before any outreach, every agent checks current customers, open deals, active sequences and every other agent's tab. Anything uncertain is skipped, not guessed.
Rule 3 · One shared record
One tab per outreach agent, one row per person, so leadership can see every touch without opening HubSpot.
Rule 4 · One Chrome job at a time
Before any job in Chrome, an agent waits until a running job from another agent posts that it's done, for up to 3 hours, and says so in one line while it waits.
Built-in safeguards
The agents do the research, writing, checking and record keeping on their own. Anything that speaks for GoDataFeed or changes a record waits for a person.
Keeping watch
The main chat checks the agents every day and audits the whole week, so a missed run or a broken step gets caught.
8:41am · Weekdays
Confirms every agent's morning run happened and asks for a make up run if one didn't.
9:00am · Weekdays
Benjamin's brief arrives after the agents have checked in, so it reflects today, not yesterday.
5:23pm · Weekdays
Refreshes the team overview and accomplishments, checks every agent reported, and confirms all five afternoon check-ins ran, since quiet ones post nothing.
8:00am · Saturdays
Reviews the full week: every run fired, every batch and sync done, LinkedIn logged on both tabs, drafts and replies handled, and totals that add up.
Part of Sales TaskForce. We built these agents to support GoDataFeed's sales team. Each agent adds its own numbers here every weekday by 5 PM ET; Instantly writes with its weekday sync at 3:32pm.
Accomplishments · since Nov 2, 2026
Everything the agent team has done since it went live, counted from what each agent reports. Every number below is something we can point to.
Customers and revenue
Counted where an agent sourced the lead, truly assisted, or (for Pro Golf Discount) supported Benjamin's own deal, with research, recommendations or reminders that moved the deal. A single drafted email doesn't count, and money counts only once it is signed or charged, never when it is quoted.
Converted to customer
—clients who became customers with real agent help
One time revenue
—charged up front, such as a managed build
Monthly recurring revenue
—/moper month, from each customer's plan in HubSpot, or the trial plan's list price until HubSpot shows one
Featured wins
Newest first. Each card says plainly how the agents were involved.
Protection
Some of the most valuable work is what the agents stopped.
Scorecards
Part of Sales TaskForce. Each agent adds its own numbers here every weekday by 5 PM ET; Instantly writes with its weekday sync at 3:32pm.
Running weekly since Nov 2, 2026
We built this agent to find up to 60 new ecommerce merchants and agencies every week, write each one a personal email series, and keep every reply moving until it becomes a conversation or is closed out cleanly.
Last report —
What it does
Up to 60 new prospects each Monday: 50 merchants and 10 agencies in the US and Canada, favoring companies with a timing signal.
A four email series for each person, built around what is happening at their company, in GoDataFeed's voice.
Runs the series through HubSpot after one yes from Benjamin, about 12 new starts each weekday at 9am ET.
Checks replies twice a day, updates HubSpot, drafts answers for Benjamin and closes out leads that stay quiet.
Pipeline
Each glowing dot is a lead moving through the agent's workflow, and a win bursts in color at the end. Circling stages count leads sitting there now; the others count everyone who reached that point. Point at any stage for where its number comes from.
Timing signals
A timing signal is a public sign that a company is changing how it sells online right now. That is when product feeds break, get rebuilt or get a new owner, so a note from us is useful instead of random. Every Monday the agent ranks its candidates by these signals and opens each first email with the one it found.
Looks like: A brand posts a job for an Ecommerce Director, a Marketplace Manager or a paid search lead.
Why it matters: The person who will own the catalog and its feeds is about to start, and new owners pick their tools early.
Looks like: Someone new has taken over ecommerce or digital in the last few months.
Why it matters: New leaders review the tools they inherited in their first quarter.
Looks like: A store moves from Magento or BigCommerce to Shopify, or the other way.
Why it matters: Every product feed has to be rebuilt after a platform move.
Looks like: A brand opens a TikTok Shop, a Walmart Marketplace store or an Amazon storefront.
Why it matters: Each new channel needs its own feed format and its own rules.
Looks like: A new product line, or a big jump in how many products the site sells.
Why it matters: More products means more feed errors and more fixing by hand.
Looks like: An agency hires paid media or shopping specialists.
Why it matters: A growing agency needs one feed tool it can run across many clients.
A company with no signal can still make the list when it fits well, but it ranks lower. The agent picks the right title at each company and never contacts more than two people there.
In practice
An illustration of how a single lead moves through the week.
In Clay it finds a Shopify outdoor gear brand that just opened a TikTok Shop and is hiring a Marketplace Manager. It picks their Director of Ecommerce.
The email checks out as verified, HubSpot has no history with the company and it is not a customer, so the lead is cleared. A customer match would be held back instead.
Email 1 opens with the TikTok Shop launch and how we build that channel's feed from the same Shopify catalog. Three follow ups and a short LinkedIn note are written too, and a review pass removes any claim the record can't back up.
Benjamin reads the whole batch in the morning check-in and says yes, or pulls anyone out first.
The contact is created in HubSpot and email 1 goes out from Benjamin's mailbox. Follow ups land every three business days.
The LinkedIn Connect agent sends a connect request with the short note, after Benjamin's Go for the week.
She replies "let's find a time". The agent updates HubSpot, creates a task and drafts the answer in Outlook, and Benjamin sends it. If nobody replies after the last email, the lead is closed out cleanly.
A real one: a Cold Outbound lead accepted the LinkedIn connect on Nov 13 and asked for time. That became the Montana Knife Company call Benjamin held on Nov 16.
How it works
The agent does itWaits for Benjamin
Searches Clay for merchants and agencies with a reason to talk now.
AgentVerifies every email and skips anyone we already know, customers included.
AgentA four email series per person, reviewed for any claim the record can't back up.
AgentThe full batch arrives Monday morning. One yes from Benjamin.
BenjaminContacts go into HubSpot and start the series, about 12 a weekday.
AgentReads replies twice a day and drafts answers for Benjamin to send.
AgentA week with this agent
All times Eastern. Weekdays only. Every weekday it also adds its numbers to this dashboard by 5 PM.
Finds new prospects in Clay, verifies, checks HubSpot and writes every email.
Reads HubSpot and Outlook, updates statuses, drafts any replies.
What came in overnight. Mondays it carries the batch for one yes.
About 12 new people start, earlier batches get their next email.
A look back at the week and a check that every rule held.
Second pass on replies.
Posts only when a reply came in, something needs Benjamin or a step failed.
Built-in safeguards
Common questions
Every company is checked against HubSpot, including current and former customers, before any email is written. Matches are held back for Benjamin to review. In the Nov 16 batch, 4 were held back this way.
Clay, a data platform we use to search companies and people and to find work emails. Only verified addresses are used, and catch all domains are skipped.
HubSpot sends them from Benjamin's mailbox as a sequence, only after his yes. Replies come back to him, and the agent drafts the answers for him to send.
Most of the first three batches are still mid series. The first monthly review on Dec 2 sets the baseline.
Results so far
Running totals, updated by the agent each weekday. Nothing here is estimated.
New people it reached personally, by week.
It is early. Most of the first three batches are still mid series, so we are not quoting a reply rate yet. The first monthly review on Dec 2 sets the baseline.
It writes every prospect to the shared sheet and to HubSpot, so the HubSpot Lead and LinkedIn Connect agents see the same history. It works alongside the Sales Nav and Instantly agents without blocking them.
Live since Nov 4, 2026
We built this agent to reach out only when there is a real reason the timing is right: a company studying our pricing, a team hiring for a competitor's tool, or a past deal worth a second look.
Last report —
What it does
Flags companies visiting our pricing, demo and comparison pages. Blog reads and email opens don't count.
Spots merchants hiring for Feedonomics or DataFeedWatch skills, a public sign they run feeds on a rival tool.
Rereads lost deals and old demos, learns why they stalled, and writes back only when something real has changed.
Updates HubSpot, creates the follow up task and drafts the answer in Outlook for Benjamin to send.
Pipeline
Each glowing dot is a lead moving through the agent's workflow, and a win bursts in color at the end. Circling stages count leads sitting there now; the others count everyone who reached that point. Point at any stage for where its number comes from.
Three lanes
Each lane has its own trigger, its own HubSpot sequence and its own lead source, so every result traces back to the lane that started it. Every email address is verified before a lead is shown, and each new contact gets a HubSpot note saying why it was added.
Lane 1 · Buyer intent
Example: A retailer looked at our pricing page several times and read our Feedonomics comparison. The agent wrote to their ecommerce lead, and Heartland America booked a demo for Nov 19. The Blessed Performance win also traced back to a Warm Signal.
Lane 2 · On a rival tool
Example: A company running three online storefronts posted a marketplace role asking for Feedonomics experience. The agent found the person who owns their feeds and asked whether keeping product options consistent across all three stores is handled in house today.
Lane 3 · Win back
Example: A demo from the summer stalled. The agent reads the meeting notes, finds what held the deal up, and opens with the one thing that has changed since, so the email answers the reason it stalled.
In practice
How a buyer intent lead moved from a page visit to a booked demo.
HubSpot shows a retailer visited our pricing and comparison pages three times this week. A university that showed intent the same day is dropped, because it doesn't sell online.
Not a customer, not in any sequence, and nobody on our team or any agent contacted them in the last 90 days, so the company is cleared.
It finds the person who owns product feeds, verifies the email, writes four emails around their situation and adds a HubSpot note on why they were picked.
The batch arrives in the morning check-in with every email. Benjamin approves, edits or holds it.
The lead is enrolled in the Warm Signal sequence in HubSpot, paced to protect the shared mailbox, and logged on its tab. The LinkedIn Connect agent adds a connect in Friday's batch.
"Can we see a demo?" The agent updates HubSpot, creates a task and drafts the answer in Outlook. Benjamin sends it, and the demo goes on the calendar.
How it works
The agent does itWaits for Benjamin
Checks buyer intent, job posts and past deals for a real reason to reach out.
AgentDrops customers, anyone in a sequence or contacted in 90 days, and bounced addresses.
AgentFinds the person who owns product feeds and writes for their situation.
AgentThe batch arrives with every email. Benjamin approves, edits or holds it.
BenjaminEnrolls each person in the lane's own HubSpot sequence and logs them.
AgentUpdates HubSpot and drafts the answer. Benjamin sends it.
BenjaminA week with this agent
All times Eastern. Weekdays only. Every weekday it also adds its numbers to this dashboard by 5 PM.
Finds teams hiring for a competitor's feed tool.
Lost deals and old demos worth a win back, for one yes.
Reads buyer intent in HubSpot. Stays quiet when nothing real shows up.
Confirms every morning run happened and what it found.
Routines, credits, junk companies and the tracker.
Replies since the morning, including LinkedIn replies on its own tab. Posts nothing on a quiet day.
Built-in safeguards
Common questions
The email never mentions the visit. It writes about their situation and the problem we solve, the same way a well timed note from a sales rep would.
On purpose. It only writes when there is a real reason. A quiet morning means nothing passed the screen, not that it didn't run.
No. Anyone already in a sequence, or contacted by our team or any agent in the last 90 days, is skipped. It also stays clear of the cold outbound agent's contacts.
Results so far
Running totals, updated by the agent each weekday. Nothing here is estimated.
New people it reached personally, by week.
Still early. Most follow up emails are still ahead, running through late November. The Second Look pool of about 110 leads is worked through by mid December.
It logs every touch on its own tab of the shared sheet, stays clear of the cold outbound agent's contacts, leaves Sales Navigator to the Sales Nav agent, and checks Instantly history before anyone is batched.
Running since Nov 9, 2026
Owns GoDataFeed's high volume cold email in Instantly. It reviews every campaign, plans the next one around a real reason to reach out, keeps a full record for leadership and audits itself every month.
Last report —
What it does
Reads every campaign's results and the replies themselves to learn what landed.
Picks companies with a reason to talk now and writes about their own products.
Logs every lead and reply so leadership can review it and nobody is emailed twice.
Once a month it measures what worked and proposes up to five changes.
Pipeline
Each glowing dot is a lead moving through the agent's workflow, and a win bursts in color at the end. Circling stages count leads sitting there now; the others count everyone who reached that point. Point at any stage for where its number comes from.
Counts from the live Outdoor campaign in Instantly. Every lead is also tracked on the instantly tab of the shared sheet.
In practice
How the live campaign was built and how it is watched.
Benjamin chose outdoor and sporting goods brands as the industry.
The agent built a list of outdoor brands with a real reason to reach out, and wrote about their own products, not generic claims.
Every company was checked against customers, open sequences and the other agents' tabs, so nobody is emailed twice.
Benjamin read sample emails before anything was built.
The emails come in two short tracks: a real finding about their listings, or a note on their current setup. Benjamin uploaded and launched the campaign on Nov 9.
Each sync reads the live campaign in full and only new replies on finished ones, then updates the shared tracker and its numbers here.
It found 79 former customers in the campaign. On Benjamin's word they were marked do not contact, and the next sync confirms none got another email.
On the last weekday of the month it reviews what worked and proposes up to five changes for Benjamin to pick.
How it works
The agent does itWaits for Benjamin
Benjamin chooses the next vertical.
BenjaminBuilds the list around a real reason to reach out.
AgentAgainst customers, sequences and every other agent's work.
AgentBenjamin reviews a sample before anything is built.
BenjaminShort emails, two tracks: a real finding or their current setup.
AgentBenjamin launches. The agent tracks every reply.
BenjaminA week with this agent
All times Eastern. Weekdays only. Every weekday it also adds its numbers to this dashboard by 5 PM.
Reads campaigns still sending in full and only new replies on finished ones, then updates the shared record and its numbers here.
Last weekday of the month: what worked, with up to five changes for Benjamin to pick.
Built-in safeguards
Common questions
Instantly runs higher volume campaigns from separate sending domains. Cold Outbound sends a small, personal weekly batch from Benjamin's own mailbox through HubSpot.
Benjamin. The agent prepares the list and the emails, but he uploads and launches, and it never edits a live campaign or replies to a lead.
Every new list is screened against customers, sequences and the other agents' work. Former customers found later are marked do not contact.
Results so far
Running totals, updated by the agent with its Monday, Wednesday and Friday syncs. Nothing here is estimated.
New people the live campaign emailed, by week.
Still early: the Outdoor campaign launched Nov 9 and has no interested replies yet. The first monthly audit on Nov 30 will judge it properly. Next up is apparel and footwear, planned and on hold until Benjamin says go.
It shares the Claude Duet Agents sheet with the other agents, screens against the Warm Signal and Cold Outbound lists, and hands qualified leads to the HubSpot Lead agent.
Live since Nov 2, 2026
An AI agent we built that keeps GoDataFeed's sales qualified leads moving in HubSpot. It finds what has stalled or been left half done, and brings back a ready fix for a one tap yes.
Last report —
What it does
Checks every sales lead each weekday morning: status, tasks, meetings, sequences and recent email.
Spots stalled leads, half finished close outs, missing follow ups and leads that already became customers.
Proposes the fix, then applies it on a yes: tasks, statuses, close outs, sequences and follow up drafts.
Researches every new lead that books a demo and leaves a short company note in HubSpot before the call.
Pipeline
The agent checks each regular SQL against HubSpot and Outlook every weekday, flags any record that no longer matches, and fixes it on Benjamin's yes.
Counts from HubSpot and the agent's own reports. This agent keeps no tab on the shared sheet; it works in HubSpot and Outlook.
In practice
Real kinds of fixes it proposes, each applied only after Benjamin's yes.
The last email went unanswered for over a week and nothing is scheduled. Fix: a follow up draft saved in Outlook and a new task.
A lead booked a meeting, but an old contact form task is still open. Fix: complete the task. Since Nov 17 the sweep flags these on any lead who booked or replied, however old.
Any of Benjamin's tasks more than 3 days overdue is flagged with a suggested next step.
A lead in the sales pipeline signs up. Fix: move it to Customer and close its open tasks.
No response after the full sequence. Fix: close it out as inactive with the reason, plus a task to try again in a few months.
A demo gets booked. It researches the company and leaves a short note in HubSpot before the call.
How it works
The agent does itWaits for Benjamin
The sweep starts on its own at 8:10 every weekday.
AgentPulls each sales lead from HubSpot and what changed since the last run.
AgentSearches Outlook and forwarded LinkedIn messages before calling a lead quiet.
AgentLists each lead that needs a fix, with the fix it suggests.
AgentBenjamin says yes, changes it, or skips it. One reply can cover several leads.
BenjaminUpdates HubSpot, saves drafts to Outlook, and reads every change back.
AgentA week with this agent
All times Eastern. Weekdays only. Every weekday it also adds its numbers to this dashboard by 5 PM.
Checks every sales lead and posts the morning report with suggested fixes.
Confirms the sweep ran. If not, a make-up run starts right away.
New replies and booked meetings since the morning, plus tasks due today. Also posts its numbers here, and stays quiet when nothing needs Benjamin.
Confirms the sweep ran every weekday and every change was made correctly.
Built-in safeguards
Common questions
No. It proposes each fix, and only applies it after Benjamin's yes. Then it reads every change back to confirm it saved.
The card now counts regular SQLs being worked. The 88 included leads it no longer tracks. No leads were lost.
Benjamin. Drafts are saved to his Outlook and he sends them.
Results so far
Running totals, updated by the agent each weekday. Nothing here is estimated.
Still early: under three weeks of data. The cleanup left 88 real leads by Nov 3, and the list now also holds every prospect the outreach agents enroll. Results on deals closed will take longer to show.
The outreach agents bring leads in. This one makes sure none of them stall once they are in the sales pipeline. It works from HubSpot and Outlook only, so it keeps no tab on the shared sheet.
Live since Nov 2, 2026
It does the "connect on LinkedIn" step for every agent's leads (HubSpot sequences, Cold Outbound and Warm Signal) from Benjamin's own account, then watches LinkedIn so no accept or reply gets missed.
Last report —
What it does
Sends each week's connect requests for every agent's leads, with a short note only when we haven't talked yet.
Reads LinkedIn alert emails twice every weekday and logs each accept and reply on the shared sheet.
Opens LinkedIn itself on Mondays and Wednesdays to catch messages, accept invites and record profile stats.
Turns inbound messages into HubSpot leads and drafts the reply, for Benjamin to send.
Pipeline
Each glowing dot is a lead moving through the agent's workflow, and a win bursts in color at the end. Circling stages count leads sitting there now; the others count everyone who reached that point. Point at any stage for where its number comes from.
Row by row on the hubspot-linkedin tab of the shared sheet, with a Source column naming the agent that found each lead.
In practice
How one week of connect requests goes out and comes back.
It gathers everyone due a connect: the week's Cold Outbound batch, Warm Signal leads from all three lanes and LinkedIn steps due in HubSpot sequences. A backstop adds any lead from those lanes still missing a connect.
Customer matches, people with no HubSpot record and anyone it can't find are held back, and the post says how many.
A short personal note goes only to people we haven't talked to yet.
Benjamin taps Go once for the whole week.
It waits for any other Chrome job to finish, then sends in two halves with a pause, posts a done line and closes the HubSpot tasks. Invites still pending after three weeks are withdrawn.
It reads LinkedIn alert emails and logs each accept or reply on its own tab and on the source agent's tab. That agent follows up from its own tab.
On a tap, it opens LinkedIn to accept invites from real people, decline newsletter invites and surface new messages.
A real one: a Cold Outbound lead accepted on Nov 13 and asked for time. That became the Montana Knife Company call on Nov 16.
How it works
The agent does itWaits for Benjamin
Gathers the week's LinkedIn steps from HubSpot sequences and the Cold Outbound and Warm Signal leads.
AgentA short note only for people we haven't talked to yet.
AgentThe plan posts with each person and note. One Go covers the week.
BenjaminSends each request from Benjamin's own LinkedIn and closes the HubSpot task.
AgentMarks every accept and reply on the shared sheet, twice a day.
AgentNew conversations get a drafted reply that Benjamin sends.
BenjaminA week with this agent
All times Eastern. Weekdays only. Every weekday it also adds its numbers to this dashboard by 5 PM.
Reads LinkedIn alert emails, updates the sheet, posts what it found.
Accepts invites from people, declines newsletter invites, records stats, surfaces messages.
One combined batch for every agent's leads, plus a backstop for any Cold Outbound, Warm Signal, Competitor Displacement or Second Look lead still missing a connect. Waits for one Go, then sends in two halves.
Reads alerts since the morning. Posts only when there is a new accept, a reply or a failure.
Built-in safeguards
Common questions
It sends connect requests with a short note. It never sends follow up messages on its own. Any reply in a conversation is drafted for Benjamin.
Everything goes out from Benjamin's one LinkedIn account, so one agent keeps the pace safe, avoids double requests and keeps one record.
It logs the reply on both tabs, and the agent that found the lead checks its own tab twice a day and follows up.
Results so far
Running totals, updated by the agent each weekday. Nothing here is estimated.
People no other agent had reached, by week, so the team total counts each person once. Every connect request it sent, for all agents' leads, is in the numbers above.
Still early. The first accepts, all on Cold Outbound leads and credited to that agent, came in from Nov 8, and one led to a call on Nov 16. Most requests went out between Nov 2 and Nov 13, so more answers are still coming.
It keeps its own tab on the shared Claude Duet Agents sheet, credits other agents' leads to their tabs, and passes any reply from their prospects straight to them.