Agentic Platform | Best Practices in an AI Centric World

Last updated: February 3, 2026

How to align your PSA with learning AI

zofiQ is not a traditional workflow automation tool. Classic automation is rule-based. You tell the system exactly what to do (for example, “if X, then Y”), and it repeats that logic forever. zofiQ works differently. It learns from your historical tickets, applies those learned patterns to new tickets, and improves when technicians correct it.

That learning model is powerful, but it comes with a hard truth:

zofiQ will learn whatever your PSA has been teaching your team, good or bad.

If your environment is consistent, zofiQ learns clean patterns and becomes highly accurate. If your environment is inconsistent, zofiQ learns the inconsistency. The most common causes are structural issues inside the PSA:

  • Too many service boards that overlap in purpose

  • Statuses that are vague, duplicated, or interpreted differently by different people

  • Type/Subtype/Item lists that are bloated, redundant, or used inconsistently

  • Missing or incomplete ticket fields that force people to guess during triage

When those conditions exist, the same type of issue (like password resets, new user setups, printer problems, or M365 access) can end up being categorized multiple ways depending on who handled it, what board it landed on, or what data was missing at the time. From an AI perspective, that means there is no single correct pattern. There are multiple competing truths in the history. The result is predictable: accuracy stalls, technicians feel like they have to babysit the system, and trust erodes.

The goal is not to create arbitrary rules for the sake of the AI.

The goal is to make your PSA easier for humans to operate and easier for zofiQ to learn from. When you simplify and standardize the structure of your PSA, you create one clear path for common issues. That is when zofiQ shifts from interesting to reliable.

The four best practices below are the foundation for getting there:

Best Practice 1: Minimize Your Service Boards

Service boards should represent materially different workflows or ownership. If two boards share the same team and lifecycle, consolidate them into one board with views/filters.

Why this matters for zofiQ:

Every board is a routing decision zofiQ needs to make. 12 boards = 12 ways to route incorrectly. 3 boards = 3 ways to route incorrectly. The math is simple.

Target: 

3-5 service boards maximum

Action:

  • Run a usage report: Which boards received tickets in the last 90 days?

  • Deactivate boards with <10 tickets/month

  • Consolidate boards where the same team handles both

  • Use board filters and saved views instead of creating new boards

Best Practice 2: Keep Board Statuses Simple and Clear

ConnectWise guidance:

Each board should have a clear lifecycle with unambiguous statuses. Status names should indicate what they mean. One closure path and avoid multiple "done" statuses.

Why this matters for zofiQ:

zofiQ learns status progressions from historical data. If your board has 15 statuses with ambiguous names, zofiQ can't learn reliable patterns about when to change status.

Common problems:

  • "Pending" (pending what? customer? approval? parts?)

  • "Waiting" (same problem as Pending)

  • Multiple closed statuses: Closed, Resolved, Completed, Fixed

  • Status names that mean different things on different boards

Target: 

Under 6 statuses per board

Action:

  • Rename ambiguous statuses: "Pending" → "Pending Customer Response"

  • Consolidate near-duplicates: Keep "Waiting on Customer," delete "Pending"

  • Have ONE closed status per board

  • Deactivate unused statuses

Best Practice 3: Start Small With Type/Subtype/Item

ConnectWise guidance:

"Start small, start somewhere, and let your technicians guide you." Build your categorization structure around how your team actually thinks about issues, not how you think they should think about issues.

Why this matters for zofiQ:

zofiQ learns categorization patterns from historical data. If you have 200 categories and technicians pick randomly, zofiQ learns chaos. If you have 30-50 clear categories that technicians use consistently, zofiQ learns accurately.

Common problems:

  • Too many options (183 types/subtypes/items across all boards)

  • Categories named "Other," "Misc," "General" that become dumping grounds

  • Overlapping categories where multiple options could be correct

  • Categories that haven't been used in months but are still active

Target: 

Under 50 total Type/Subtype/Item combinations that cover 80% of your tickets

Action:

  • Run a report: What Type/Subtype/Item combinations were used in the last 90 days?

  • Deactivate anything used <5 times in 90 days

  • Eliminate "Other/Misc/General" categories

  • Consolidate overlapping categories

  • Create 10-15 "golden ticket" examples showing correct categorization for common issues

Best Practice 4: Enforce Complete Ticket Data

Every ticket should have Type, Subtype, Item, Priority, and meaningful resolution notes. Create reports that show incomplete tickets and hold technicians accountable.

Why this matters for zofiQ:

Incomplete tickets are useless training data. If 30% of your historical tickets have no categorization, zofiQ can only learn from the other 70%—and even those might be inconsistent.

Common problems:

  • Resolution notes: "fixed," "resolved," "done" (no detail)

  • Tickets closed without Type/Subtype/Item

  • Bulk status changes (cleanup that corrupts learning)

  • Merged tickets without documentation

Action:

  • Make Type/Subtype/Item required before ticket closure

  • Require meaningful resolution notes (use templates)

  • Stop bulk-editing tickets as cleanup—it poisons AI training data

  • Create a report of tickets missing categorization and assign cleanup to specific people