Projects & agent library

Practical systems, with their real status.

Three featured projects, an interactive demo, and a compact library of 20 agents. These are my own builds and demonstrations, not client results.

Status labels:PrototypeIn DevelopmentBuilt & TestedClient-Deployed

Discuss a Project Try the demo

Featured projects

Three builds in detail.

CRM Data Quality Auditor

Built & Tested

Business problem: Teams cannot tell how reliable their CRM data is, so reporting, routing and forecasting all rest on records nobody has checked.
What it does: Reads CRM data, checks fields, completeness, associations, duplicates and source tracking against thresholds, and returns findings ranked by severity. Each finding has an observation, a business impact, a recommended fix and an effort estimate.
Inputs and outputs: In: HubSpot records and properties. Out: a severity-ranked findings report you can turn into a fix roadmap.
Tools: Python, HubSpot API.

Output format · illustrative, not client dataSeverity: High Finding: Contacts with no company association Observation: [count and share of contacts affected] Impact: Account-level reporting breaks Recommendation: Associate or enrich the unlinked contacts Effort: [estimate]

Governance: Read-only audit. It reports; it does not change records.
Status and limits: Built and tested as an independent project. Not currently client-deployed. Thresholds are defaults that should be agreed with each team.

Discuss a similar build

Buying Signal Detection Agent

Built & Tested

Business problem: Relevant company events, such as hiring activity, are noticed late or not at all, and reps cannot see why an account matters now.
What it does: Detects buying signals for target accounts and attaches the evidence for each one, so a signal can be checked rather than trusted blindly.
Inputs and outputs: In: your target account list and the signal types you care about. Out: signals with source evidence, ready to review or route to sales.

Governance: Every signal carries its evidence. Anything that would write to the CRM goes through validation and, where you require it, approval.
Status and limits: Built and tested as an independent project. Not currently client-deployed. Build documentation is available on request.

Discuss a similar build

AI Lead Intelligence for HubSpot

Prototype

Business problem: Buying signals sit in separate tools, so sales has no clear order of priority and no evidence trail behind it.
What it does: Combines CRM, engagement, intent and product signals into an explainable account score. A language model drafts a brief from the evidence, with hypotheses labelled. Proposed HubSpot updates wait for approval.
Inputs and outputs: In: CRM and account data plus whichever signal sources you connect. Out: a score with itemised contributions, a brief, and proposed property updates.

Governance: Validation before writes, human approval for sensitive fields, retries on errors, an audit log, and no lifecycle stage changed by a score alone.
Status and limits: Prototype. The demo below runs on fictional accounts with simulated integrations. It is not connected to a live HubSpot portal, and the scoring has not been validated against historical outcomes.
Architecture: See the eight-stage reference design.

Try the interactive demo Discuss a similar build

Interactive demonstration

See one account go from evidence to an approved CRM update.

About a minute and a half. The scores are calculated live from the evidence on screen, and every record, person and approval is fictional and simulated.

Pick an account
Check the evidence and how the score is built
Read the brief and recommended action
Approve or reject the proposed updates

Explore three fictional accounts to see how buying signals, missing data and customer expansion activity affect prioritization, generate sales briefs and inform proposed CRM updates.

Weights, thresholds and approval rules shown are examples. In a real project they are agreed with your team.

Agent library

Twenty agents across the revenue workflow.

Each card shows the problem, what the agent does, and its status. Detailed pages exist only for the featured projects above. Ask me to walk through any of the others.

20 agents

CRM & RevOps Intelligence

CRM Data Quality Auditor

Nobody knows how reliable the CRM data really is.

Audits fields, completeness, associations, duplicates and source tracking, then returns severity-ranked findings with impact and recommended fixes.

Built & Tested

CRM & RevOps Intelligence

Duplicate Detection & Merge Agent

Duplicate records double-count contacts and send reps after the same lead twice.

Finds likely duplicate records and recommends or manages their resolution under approval rules you set.

Built & Tested

CRM & RevOps Intelligence

Contact–Company Association Agent

Contacts without the right company break account-level reporting.

Detects missing or incorrect contact-to-company relationships and proposes corrections.

Built & Tested

CRM & RevOps Intelligence

Property Normalization Agent

Inconsistent property values make segmentation and reporting unreliable.

Standardizes inconsistent property values and data formats against rules you define.

Built & Tested

CRM & RevOps Intelligence

Lifecycle Stage Auditor

Records sit in the wrong lifecycle stage, so funnel numbers mislead.

Identifies incorrect lifecycle stages and the process inconsistencies behind them.

Built & Tested

CRM & RevOps Intelligence

Sales Process Compliance Auditor

Deals skip steps and required data is missing.

Checks whether CRM records and deals follow your defined sales process and flags the gaps.

Built & Tested

CRM & RevOps Intelligence

RevOps Copilot

Routine RevOps questions and CRM chores depend on one person.

A conversational interface for RevOps questions that can run approved CRM actions.

Built & Tested

Account Intelligence

Lead Qualification Agent

Qualification is inconsistent from rep to rep.

Evaluates leads against your written qualification criteria and shows which criteria were met.

Built & Tested

Account Intelligence

Account Research Agent

Manual research eats selling time.

Gathers and summarizes relevant company intelligence with the sources attached.

Built & Tested

Account Intelligence

Stakeholder Mapping Agent

Reps miss key people in the buying committee.

Identifies relevant stakeholders and buying-committee relationships for an account.

Built & Tested

Signal Intelligence

Buying Signal Detection Agent

Relevant company events are noticed late or not at all.

Identifies buying signals, such as hiring activity, and attaches the evidence for each one.

Built & Tested

Signal Intelligence

ICP Drift Detector

Your best customers may no longer match your stated ideal customer profile.

Detects when the characteristics of converting customers diverge from the intended ICP.

Built & Tested

Sales & Pipeline Intelligence

Lead Routing Agent

Leads wait or go to the wrong owner.

Assigns leads according to defined rules and business context.

Built & Tested

Sales & Pipeline Intelligence

Lead Prioritization Agent

Sales sees a long list with no clear order.

Ranks leads by fit, signals and sales priority with an explainable score.

Built & Tested

Sales & Pipeline Intelligence

Stale Opportunity Rescue Agent

Open deals go quiet without anyone noticing.

Identifies stalled opportunities and recommends follow-up actions.

Built & Tested

Sales & Pipeline Intelligence

Deal Risk Detection Agent

Risk shows up only when the deal is already lost.

Surfaces deal-level risks that may affect progression or closure.

Built & Tested

Sales & Pipeline Intelligence

Revenue Forecast Risk Engine

Forecasts rest on records that do not hold up.

Identifies risks and inconsistencies in the data that could affect the forecast.

Built & Tested

Marketing & Revenue Intelligence

Attribution Anomaly Detector

Attribution gaps hide which channels work.

Identifies suspicious, missing or inconsistent attribution data.

Built & Tested

Marketing & Revenue Intelligence

Campaign-to-Revenue Analyzer

Campaign reports stop at leads and never reach revenue.

Connects campaign performance with pipeline and revenue outcomes.

Built & Tested

Marketing & Revenue Intelligence

GTM Experiment Analyzer

Experiment results are hard to read and easy to over-claim.

Evaluates GTM experiments and helps interpret their results.

Built & Tested

All agents shown are built and tested by me as independent projects. None is currently listed as Client-Deployed. A label changes only when it becomes true.

Next step

Need one of these built for your stack?

Tell me which problem you want solved. I will tell you what a first scoped version would involve.

Discuss a Project Explore My Work