More outbound volume is hiding weaker work
This week’s signals favour tighter exclusions, shorter sequences, and lower AI costs.
By Dimitar Petkov, LeadHaste
Subscribers got this on 4 Sept 2026. Get the next one a week before it goes up here.
We think the next improvement in outbound will come from subtraction, not another layer of automation. This week's signals point in the same direction: one live test removed 106 prospects before outreach, shorter sequences outperformed longer ones, and agentic SDR pilots traded a 6.4x increase in sends for a decline in positive reply rate. Meanwhile, cheaper cached AI context makes enrichment easier to afford, not automatically more useful. The common thread is not less technology. It is a better decision about where technology belongs. A system that cannot say no to a poor-fit account will make its reporting look busy while the work underneath gets less precise.
The tempting response is to use the cost saving to process a bigger list. We would not. Put it into tighter fit rules, a clearer point of view, and a sequence people can actually maintain. Review the accounts that reply, the accounts that decline, and the accounts that were excluded. Then use that evidence to adjust the next run. That is how an outbound system compounds: each round improves the inputs instead of merely increasing activity.
Start with the exclusions
A practitioner test excluded 106 records from a raw prospect pool before sending, based on company size, missing decision-makers, and poor fit. The remaining 26 prospects received enrichment and pre-warm LinkedIn touches, producing a 26% reply rate and four booked calls. It is a small, self-reported test, but the order of work is the useful part: decide who should not enter the system before spending effort on personalization. That decision also makes a weak result easier to diagnose, because the team can inspect fit criteria before changing copy or volume. Use a short exclusion review before each list release, and record the reason each account was removed so the rule can be tested in the next campaign. For an outbound operator, tighter exclusion rules can protect sender capacity for accounts that have a credible reason to reply.
Source ↗Cache context, not judgment
Anthropic's Fable 5.1 reportedly cuts cached-context reads by 75%, to $0.25 per million tokens, while headline pricing stays flat. That changes the cost of repeatedly supplying an ICP and scoring context to AI enrichment across a large account set. The saving is real only when the repeated context is already sound, because a cheaper bad fit is still a bad fit. Teams should separate reusable account rules from the evidence that genuinely changes by company, then review whether the output is changing a commercial decision. That creates a cleaner comparison between research cost, decision quality, and the amount of manual review a system still requires before a message is approved for sending. For a buyer of outbound, lower AI enrichment costs should fund better account research and fit checks before they fund more volume.
Source ↗Keep the sequence short
Shared same-account tests put three-message sequences at a 9.8% reply rate, while sequences with five or more messages underperformed a single message. The same discussion says AI-personalized copy trailed human-written templates by 12% in those accounts. These are reports from GTM engineering circles, not a universal benchmark, but they support a useful operating rule: prove each message earns its place. A sequence should have a reason for the first contact, a different reason for the follow-up, and a clear stop condition. That rule also makes testing cleaner: change one variable, watch the reply quality, and keep the follow-up only when it adds a reason to respond without exhausting sender reputation or the team's time on avoidable work each week. For a team running outbound, cutting unproductive follow-ups can preserve attention for stronger templates and faster reply handling.
Source ↗Volume can dilute intent
Agentic SDR pilots discussed by GTM engineers increased send volume 6.4x while positive reply rates fell from 2.1% to 1.3%. That does not make automation the problem. It makes the measurement choice the problem: an outbound system should measure whether added activity creates more qualified conversations, not just more delivered messages. The useful review is not whether the machine can send more, but whether it changes the quality of conversations that reach the sales team and produces reliable evidence for the next decision about accounts and message fit. For a sales leader buying outbound, require reporting on positive replies and qualified meetings alongside send counts before approving more automation.
Source ↗Clay Workflows
Clay is moving mature table-based GTM automations toward its newer Workflows building block, and is running live sessions to help customers convert existing Tables. Teams with established Clay tables should watch the migration closely rather than wait for a forced rebuild. Start with the tables that shape account qualification, enrichment, and routing, because those are the rules that affect daily outbound work. Treat the migration as process maintenance, not a reason to rebuild rules that already work. For an outbound operator, map the table logic you own now so a move to Workflows does not interrupt enrichment or routing.
Want this running on your own pipeline?
We build and run the outbound system behind these notes, in your name, on infrastructure you own. It starts with a free ICP and campaign-fit review.
Book your free ICP review →