Why score on three layers instead of one?

Because no single layer predicts worth touching today on its own. Fit without intent is a list of nice targets with no buying window. Intent without fit is a warm account in the wrong ICP. Engagement without fit or intent is noise. Only when fit and intent both cross a threshold does an account belong on the daily queue.

How big should the daily queue be per rep?

Size the queue to the rep's call capacity. For most SDRs, 20 to 30 accounts per day is the practical ceiling if the rep is expected to prepare and open each one. Surfacing 200 accounts produces the same outcome as surfacing none, because nothing gets real attention.

How do we avoid the agent obsessing over recent signals?

Keep fit scores separate from intent. Fit is slow-moving and does not deserve time-decay. Intent is fast-moving and does. When the two are combined into a composite with clear weights, the agent cannot drift into a pattern of chasing the latest signal at the expense of real buyer fit.

How often should we recalibrate the model?

Quarterly recalibration is the baseline. Pull closed-won and closed-lost data, validate that the composite score is still tracking outcomes, and reweight the layers if the correlation has drifted. Anything more frequent tends to produce noise rather than signal.

How AI Agents Decide Which Accounts Are Worth Touching Today

An AI agent deciding who to call today runs a continuous ranking function across the account universe, evaluating fit, intent, and engagement in parallel and surfacing only the accounts where the composite score clears the action threshold within the current day. The agent is not building a static list. It is refreshing the list on every new signal.

The 3 Scoring Layers

ICP fit (firmographics, technographics, win patterns), buying intent (funding, hiring, category research), and engagement (email opens, content views, past replies). Fit and intent must both cross a threshold; engagement breaks ties.

Time-Decay on Signals

Exponential weighting pushes recent events to the top. 0 to 7 days gets full weight, 8 to 30 days partial, 31 to 90 days near zero. This is how a 65/100 account becomes 88/100 when a funding event and VP hire cluster in one week.

Tier 1 Discipline

Cap Tier 1 at 10% of total accounts. Teams that overload Tier 1 see 2.3x lower engagement. Signal-qualified leads convert 47% better and close at 43% larger deal sizes.

Build Order

Build ICP fit first, then intent, then engagement. Calibrate quarterly against closed-won data so the composite score keeps tracking outcomes as markets shift.