Who to hire for Bluebirds

Imagine your best seller opens her inbox on a Monday and finds forty new prospects, each with a short reason to call. Thirty are people she already talks to, six are at companies that asked not to be contacted, and two are in a territory that belongs to a colleague. The tool did what it was built to do. The org around it was not ready. That is the hiring problem for Bluebirds, and it is more about your records than the product.

Salesforce completed the acquisition on 13 August 2025. The announcement describes the technology this way: "Salesforce has signed a definitive agreement to acquire Bluebirds, a leading AI-powered prospecting platform that helps sales teams identify and engage high-potential leads using advanced lead intelligence and data enrichment." The same announcement said Salesforce planned to bring these capabilities into Sales Cloud and Agentforce. For a buyer, the practical reading is that prospecting signals and enrichment are heading for the same objects your sellers already use, which makes the quality of those objects the first thing to examine.

#Start from the record, not the signal

Every prospecting workflow ends in a write to Salesforce. A person or company is created or updated, an owner is assigned, a task or sequence step is scheduled, and a source is recorded. Each of those writes can go wrong in a specific way, and a competent partner can describe all of them before you ask.

Duplicates come first. Salesforce provides matching rules and duplicate rules for Leads, Contacts and Accounts, and these decide whether a new record is blocked, allowed with a warning or merged into an existing one. Rules written for manual entry often fail against automated creation, because machines produce records faster than humans and with slightly different spellings. Ask a candidate to show how a new signal is compared with existing Leads, Contacts and open Opportunities, and what happens in each outcome.

Ownership comes second. A Lead created by an automation needs an owner at creation, using the same territory and routing rules your sellers trust. If the answer is a queue that someone triages on Fridays, the prospects will have gone cold by then.

Consent comes third, and it is where the legal exposure sits. In Australia, the Spam Act 2003 requires consent for commercial electronic messages, and the Privacy Act 1988 governs how personal information is collected and used. Enrichment adds a further question, namely where the extra data came from and whether the person would expect it to be used this way. A partner should ask for your consent policy in the first meeting and propose where in the record the evidence of consent sits.

#What a partner does on a Bluebirds project

The work divides into five parts, and your proposal should name each.

Data model. Decide which fields store the signal, the source, the enrichment date and the consent status. Prefer fields that reports can use over a free-text note.

Matching and creation. Write the rules that decide whether a signal creates a Lead, updates a Contact or is discarded. Include a log of discarded signals, since they show you what the rules are doing.

Routing and assignment. Reuse your existing territory logic. Add an owner-assignment test that runs against a sample of real records.

Seller experience. Decide where the signal appears, what the seller sees as the reason to act, and how a seller marks a signal as irrelevant so the system learns from it.

Measurement. Define the comparison you will use before launch. A group of sellers or accounts that does not receive the signals is the only honest baseline.

If agents will draft or send outreach, the Agentforce part is a further piece with its own permissions, guardrails and review step. Keep it separate in the statement of work so that a delay there does not hold up the data work.

#Skills and certifications

The credential that fits the Sales Cloud side is the Salesforce Certified Agentforce Sales Consultant, which was previously the Sales Cloud Consultant. It tells you the person understands lead and opportunity management in the product. Salesforce Certified Platform Administrator covers duplicate rules, assignment and validation. Salesforce Certified Platform Data Architect matters if your Lead and Contact volumes are large or if enrichment will write to many fields. Salesforce Certified Agentforce Specialist is relevant where an agent will prepare or send messages. Integration work, if the signals arrive from outside Salesforce, calls for someone with the Platform Integration Architect or at least Platform Developer background.

Credentials only point at a role. Ask each person on the team to describe a duplicate problem they solved, and what the rule looked like in the end.

#A worked example of one signal

Follow a single signal through the design, because that is how you will judge a proposal. A signal arrives saying that a finance director at a mid-sized company has changed jobs. First the rules check whether the person already exists. Suppose she is a Contact at an Account your team services. The correct outcome is not a new Lead. It is a task for the Account owner, with the reason attached, and an update to the Contact's role if your policy allows enrichment to write to that field.

Now suppose she is unknown, but her new company is an existing customer. The outcome should be a Contact on that Account, owned by the Account owner, not a Lead sent to a different seller. Finally suppose neither she nor the company exists. A Lead is justified, but only if the source, the signal and the consent status can be recorded, and only if the territory rules assign it to a person immediately.

Three outcomes from one signal, and each needs a different rule. A proposal that describes only the third is describing a demo. When you review a design, ask the partner to produce a table with every outcome, the rule behind it and a sample of real records that landed in each. The table is also the best training aid for sellers, who will trust the output once they see how it decides.

#Seller adoption

Sellers do not read governance documents, they read their queue. If the first ten items in the queue are poor, the project is over, whatever the data model looks like. For that reason, ask the partner to plan the first two weeks after launch as a tuning period, with a short review every few days where sellers say which signals were useful and which were noise. Put an owner on the feedback, and make sure the changes reach the rules quickly. Good partners also ask for a few sellers to join design sessions early, since those people will explain in minutes what a consultant takes weeks to infer.

Be wary of measuring success only by the volume of signals created. A smaller number that sellers act on is worth more than a large number they scroll past. Report on acted-on signals, meetings booked and, much later, pipeline that came from them, always against the comparison group.

#Red flags

The first red flag is a conversion promise. Nobody can know how your market responds before a test, and a partner who quotes a rate has borrowed it from somewhere else. Ask how they would design the test, how long it must run, and what result would lead them to recommend stopping.

The second is a plan that creates Leads first and cleans them later. Cleaning afterwards means merging records that sellers already worked, and it damages trust in the data.

The third is silence about consent. If the partner does not ask where permission is recorded, they will not build for it.

The fourth is a design that bypasses your routing. A parallel route to the seller creates two sets of rules and two sets of arguments.

The fifth is a scope that includes outbound sequences, telephony and a new dialler. Those are separate decisions with their own buyers.

#Reporting you should insist on

Ask for three views before launch. One shows signals received, created, merged and discarded, by reason. One shows ownership: how quickly each new record reached a seller and how many were reassigned. One shows consent status across everything the project touched. A partner who builds these early will find problems early, and your operations lead will have something to read each week.

#Ten questions for the room

1. When a signal arrives for a person who already exists as a Contact, what happens, and which rule decides it?
2. How do you stop two signals for the same company from creating two Leads?
3. Who owns a new Lead at the moment it is created, and how is that decided?
4. Where do we record the source of each signal and the date of enrichment?
5. Where is consent stored, and how does a seller see it before acting?
6. Which signals will you recommend we ignore at the start?
7. How will a seller mark a signal as wrong, and what happens to that feedback?
8. What comparison group will we use to judge whether the signals helped?
9. Which parts of this are Sales Cloud configuration, which are integration and which are Agentforce?
10. What would make you recommend we pause the pilot?

#Checking evidence

Ask a previous client what happened to their duplicate rate in the first month, how many signals were discarded and why, and what the sellers said in week three. Those answers are specific enough to be checked.

Ask for an anonymised matching-rule design and an owner-assignment test plan. A partner who has built these will have them on file.

Ask to speak to a sales operations lead rather than a sponsor. Operations people remember what broke.

#Commercial models and the statement of work

A fixed-price design phase followed by time-and-materials build works well. The design phase ends with the data model, the matching rules, the routing design and the measurement plan, all written down. Build can then be estimated against a document rather than a hope. For what influences the quote, see the implementation cost page, and for the licence side see the licence page.

The statement of work should name the segment, the signals in scope, the objects written to, the fields added, the duplicate rules to be created or changed, the routing logic reused, the consent evidence required, the measurement design, the training for sellers, and what is out of scope. Add an acceptance test that runs a hundred real records through the creation rules and shows the outcomes in a table your sales operations lead can read.

#When to skip a partner

A small team with a tidy org, one segment and a capable administrator can run a pilot without outside help. If you have fewer than a handful of sellers, a single Lead source and clear duplicate rules already, the work is configuration. Buy help when your Lead data is untidy, when several teams create Leads, or when consent evidence is incomplete.

For further reading, see what happened to Bluebirds, integrating Bluebirds with Salesforce, should I migrate from Bluebirds and Bluebirds and Salesforce prospecting. The primary source is the Salesforce announcement, and the certification list is on Trailhead.