The best product work is not a longer feature list. It is a platform that can do more while asking less of the people using it—and one that becomes safer and more predictable as it grows.
Over the 30 days ending September 9, the Salestools team shipped 1,714 commits across our frontend and backend, touching calling, research, enrichment, LinkedIn, inbox collaboration, agent building, infrastructure, and tenant security. The number is not the story by itself. The story is what those changes make possible for revenue teams.
Here are the releases that matter most in day-to-day work.
A power dialer built for real-world calling
Calling at speed only works when control keeps pace with volume. This month we rebuilt the foundations of the power dialer around compliance, ownership, and accurate outcomes.
The dialer now evaluates calling windows in the called party's local time, applies stricter state-level windows where required, and fails closed when it cannot make a safe decision. Phone numbers are now first-class entries in do-not-call suppression, checked before every dial—including replacement calls.
We also added an abandoned-call guard, explicit recording-consent controls on every call path, caller-ID ownership checks at session start, and automatic cleanup for sessions that should no longer be active. Call history, recordings, transcripts, and analytics now share one consistent record of what happened.
The result is a dialer designed to move quickly without losing the context and guardrails that responsible teams need.
Deep Research is now a product, not a background process
Customers can now run Analyst Research directly, select the depth they need, share a live run with teammates, and export the result as Markdown, CSV, or PDF. Runs are readable across the team, and shareable URLs make it easier to move a finding into a deal review or account plan.
Research can start from a custom prompt, a saved template, or a useful default. Under the hood, company and persona research now use fair per-user scheduling, so one large workload does not crowd out everyone else.
Most importantly, research is connected to the rest of the workflow. It can inform targeting, messaging, and account decisions instead of ending as a document someone has to manually interpret.
Enrichment that understands what a business actually is
Firmographic data tells you the size and location of a company. It does not always tell you how that company goes to market.
We added business-type classification for B2B, B2C, and government organizations, including support for companies that genuinely belong to more than one category. In the initial classification run, 17,304 companies were analyzed, with multi-tag classifications applied where a single label would have been misleading.
Teams can also bring their own Lusha, ZoomInfo, and Zeliq providers. New provider-callback infrastructure verifies signed responses and waits for the result inside the enrichment run that requested it. Predicted emails are clearly labeled and remain masked where masked search rules apply.
This is the connective tissue between better data and better decisions: classify the business correctly, gather the right evidence, and make the result available at the moment an agent needs it.
Reply intelligence now carries the full conversation forward
A reply handler should not merely detect a response. It should understand the response, preserve the context, and answer in the voice the team intended.
Rich tone-of-voice settings now reach generation end to end. The reply classifier's interpretation is passed into the writer, sentiment is saved with the approval record, and reply windows are anchored to the last message sent. We also corrected the default behavior that could remove a lead from the workflow before the reply handler had a chance to act.
On LinkedIn, Sales Navigator workflows can now send InMail on the correct route, show credit balances, and respect per-seat choices about using profile-unlock credits. Profile Visitors is now a lead source monitored twice daily, and social activity for a lead—including posts and comments—can feed better-timed outreach.
WhatsApp replies now travel through the same reply orchestration layer, bringing another channel into a consistent system for understanding, routing, and response.
The inbox now works at team scope
Managers can read their team's threads, search across that scope, and reply on a teammate's conversation with attribution preserved. Access remains bounded to the right team, but the team view is no longer read-only.
That sounds like a small distinction. Operationally, it closes a major gap: a manager can step into a live conversation without copying context into another tool or losing who sent the response.
A clearer agent builder and a real marketing workspace
The cadence builder now makes the active operating mode visible: CoPilot or AutoPilot can be set per agent and overridden per lead, while the mode that will actually run stays visible in the header. There is one Start action, a readiness gate before launch, a booking gate at runtime, and deeper step statistics with searchable drill-down.
Adaptive intelligence can incorporate CRM context, enforce source health, and show a preview before a workflow goes live. The FIELD canvas has also been redesigned as a more spatial workspace, with clearer connections between steps and less visual friction.
The new marketing platform reached its first working phase as well: an in-app shell, a real-data overview, attribution APIs and views, request-level entitlement checks, and the FIELD workspace activated for marketing use cases.
Reliability and tenant boundaries strengthened across the platform
This release also includes the work users should rarely have to notice.
We tightened authentication and tenant scoping across data access, credential branches, activity deletion, lead scoring, prospect locks, caller IDs, voicemail recordings, and intelligent-ICP learning. Elasticsearch ingest is now append-only, batch updates retain their bodies, and empty arrays can no longer silently wipe stored identifiers or domains.
Infrastructure updates improve forwarded-protocol handling, metrics collection, worker recovery, and container hygiene. Several obsolete AI and endpoint paths were removed so there are fewer competing ways to do the same job.
These changes do not add another button to the interface. They make every existing button more trustworthy.
What this month adds up to
The platform is becoming more capable in a specific way: agents can act across more of the revenue workflow while teams keep stronger control over data, cost, tone, access, and compliance.
That is the direction we will keep shipping toward. More work completed end to end. Better evidence behind every decision. Clear boundaries around what an agent may do. And fewer operational gaps for your team to bridge manually.
See the complete release timeline in the Salestools changelog, or talk with our team about putting the new capabilities to work.
