Evidence-based automation projects

Real project work, with the build status shown.

These portfolio systems and active client engagements show how MannaLabs researches, specifies, builds, tests, and safeguards AI automation. Each project separates what exists now from what still needs confirmation or production setup.

Evidence over claims

Every project description is tied to an existing specification, workflow source file, configured system, or recorded test. Planned work is never presented as a result.

Build status is explicit

Working prototypes, inactive deployments, tested workflow slices, and specification-only projects are labeled separately so you can see what exists today.

Client stage stays visible

Active client work is labeled as discovery or research, not presented as a completed case study. Private data and unverified outcomes remain excluded.

Portfolio and client projects

From working prototypes to active client discovery

No invented client results, anonymous testimonials, or projected metrics. The evidence and remaining boundary are included on every project.

Portfolio demoRevenue OperationsWorking prototype - inactive

B2B SaaS Demo Request Router

An inbound demo-request system that captures each request in Airtable, enriches the company, scores fit and urgency, assigns an available sales owner, and coordinates the handoff in Slack.

Problem addressed

Small sales teams lose the highest-intent response window when demo requests wait in forms, inboxes, or static CRM queues without ownership or an SLA.

Workflow path

Airtable form
Company enrichment
Fit scoring
Round-robin routing
Slack claim
SLA escalation

Implemented evidence

Airtable intake base, routing rules, sales roster, and event log configured
Inactive n8n workflow implements intake, scoring, routing, Slack alerts, claims, SLA checks, and a daily digest
People Data Labs enrichment and CRM-ready staging logic included

Current boundary

Slack interactivity, real rep IDs, production CRM sync, activation, and a real end-to-end test remain pending.

Portfolio demoOutbound & ResearchSource-controlled prototype - inactive

AI GTM Lead Gen Engine

A signal-based outbound system that turns an ICP brief into researched and enriched accounts, grounded outreach drafts, a human approval queue, sequencer enrollment, and reply-state updates.

Problem addressed

Lean B2B teams cannot sustain account research manually, but automating generic send volume creates weak personalization and deliverability risk.

Workflow path

ICP brief
Signal sourcing
Lead enrichment
Research brief
Human approval
Send and reply sync

Implemented evidence

Five production-style n8n workflows maintained as TypeScript source
Inactive workflows cover direct enrichment, Clay enrichment, AI drafting, approved Smartlead enrollment, and reply capture
Airtable state, activity logging, approval gates, and stop-on-reply logic are designed into the system

Current boundary

Synthetic production-path testing, credential audit, deliverability monitoring, and activation are still required before live outreach.

Portfolio demoContent OperationsPrototype tested with pinned data

Social Autopilot

A creator research and scripting system that tracks content signals, scores outliers, converts the strongest opportunity into hooks and scripts, and writes the work into an Airtable content dashboard.

Problem addressed

Small teams and owner-creators often publish from guesswork because researching trends, identifying useful outliers, and turning them into on-brand scripts takes too much daily effort.

Workflow path

Channel tracking
Outlier scoring
Trend research
Hook generation
Script drafting
Airtable dashboard

Implemented evidence

Airtable content OS includes monitored channels, trend signals, posts, reports, and dashboard payloads
Inactive Strategist Research and Script Writer n8n workers are implemented
Script Writer passed pinned-data test execution and created the expected structured post record

Current boundary

Live channel collection, asset generation, editing handoff, morning email, and dashboard deployment remain unfinished.

Portfolio demoSEO & ContentBuild-ready specification

AI SEO Growth Engine

A research-to-publishing workflow designed to turn a topic into a sourced content brief, a human-reviewed draft, a claim-and-citation check, and a tracked publishing record.

Problem addressed

Lean marketing teams spend most of their content capacity on repetitive research, briefing, drafting, and source checking while still needing human editorial control.

Workflow path

Topic intake
SERP research
Content brief
Brief approval
Draft and citation check
Final approval

Implemented evidence

Research, ICP, PRD, workflow specification, implementation plan, and operating notes completed
Two mandatory human approval gates and claim-level citation checks specified
n8n build workspace and workflow PRD template prepared

Current boundary

No automation has been built or deployed. A SERP provider, SEO data provider, and target CMS must be selected first.

Client engagementSales OperationsDiscovery - pre-build

NewGen Outbound Research Assistant

A planned internal outbound assistant for NewGen that filters leads against an approved ICP, researches each company from public sources, drafts personalized LinkedIn outreach, and coordinates human review and follow-up tracking.

Problem addressed

Business-development research, qualification, message drafting, and follow-up are difficult to run consistently when the process is spread across lead tables and manual LinkedIn activity.

Workflow path

Airtable or Sheets
ICP filter
Deduplication
Public company research
Human-approved draft
Manual send and follow-up

Implemented evidence

Client context, market research, ICP, PRD, workflow specification, implementation plan, and operating notes prepared
Workflow contract covers lead intake, qualification, public-source research, drafting, approval, manual sending, and stop conditions
LinkedIn sending and profile activity remain explicitly human-operated

Current boundary

Project owner, lead source, exact ICP rules, approval channel, volume, credentials, sample data, and production activation are not yet confirmed.

Client engagementVoice & Sales AutomationResearch - scope unconfirmed

Xcent.ai Consent-Based AI Calling Agent

A researched AI voice-agent direction for Xcent.ai focused on warm leads who explicitly request a demo or callback, with live qualification, product answers from approved knowledge, booking, and human transfer.

Problem addressed

A sales calling agent must respond naturally in real time while keeping consent, product claims, call actions, escalation, recording policy, and post-call operations deterministic and auditable.

Workflow path

Consent proof
Warm call queue
AI voice conversation
Validated tool action
Book or transfer
Post-call orchestration

Implemented evidence

Public-company research, ICP analysis, voice-agent research, sales-script research, workflow specification, and implementation planning prepared
Architecture separates the low-latency voice runtime from validated business tools and n8n post-call orchestration
Recommended pilot requires AI disclosure, verbal opt-out, human transfer, and warm-lead consent

Current boundary

The first project, owner, market, call purpose, consent proof, recording policy, telephony route, baseline, success target, and production access remain unconfirmed.

Client work

Shown honestly, without turning plans into results.

NewGen and Xcent.ai are included at their current engagement stages. Their cards describe confirmed direction and completed planning work, not production outcomes or client success metrics.

Discuss your workflow