Sales follow-up agent
Prepare or execute defined follow-up actions using CRM context and approved rules.
AGENTS / 04
Deploy agents with a job description, permissions and supervision. Controlled autonomy, not unsupervised automation.
Northbound designs AI agents around narrowly defined business processes, explicit tool access, clear escalation rules and measurable outcomes so they operate as controlled systems rather than open-ended experiments.
An AI agent should know what it can do, what it cannot do, and when a human takes over.
02 / WHAT NORTHBOUND EXAMINES
Job definitionA bounded process with clear start, finish and success criteria.
Tool accessOnly the systems and actions required for the assigned work.
Authority levelRead, draft, recommend, execute or escalate permissions defined explicitly.
Context qualityReliable data and knowledge available when the agent needs to act.
ExceptionsKnown failure modes and escalation paths.
ObservabilityLogs, monitoring and review mechanisms for important actions.
03 / COMMON USE CASES
What this capability can look like in practice.
These are representative patterns, not a prescribed technology stack. The right implementation depends on the business, systems, data and risk profile.
Service triage agent
Classify requests, gather context and route issues before a human handles the exception.
Operations agent
Coordinate repetitive multi-step administrative processes across approved systems.
Research agent
Collect, compare and synthesize defined information for human review.
Knowledge agent
Retrieve relevant internal information and prepare answers grounded in approved sources.
Monitoring agent
Watch defined signals and surface exceptions requiring human attention.
04 / MEASUREMENT
Measure the change in the business, not the novelty of the technology.
Northbound establishes the baseline before implementation and tracks the operating signals that determine whether the capability is actually creating value.
COMPLETIONTasks resolved within defined scope
ESCALATIONCases correctly handed to humans
QUALITYAccuracy and correction rate
COSTOperating cost per successful task
05 / ENGAGEMENT PATH
Move from clarity to controlled execution.
The scope expands only when the evidence supports it.
BOUND
Define the job
Scope the process, tools, data, authority and human checkpoints.
TEST
Prove behaviour
Evaluate normal cases, edge cases, failures and escalation quality.
OPERATE
Monitor in production
Track performance, cost, incidents and changes over time.
06 / CONTROL POINT
Responsible transformation is part of the design.
Northbound avoids “agent for everything” designs. The more consequential the action, the more explicit the permission model, evidence trail and human oversight need to be.
See whether AI Agents belongs in your roadmap. Start with the business.
Use the Northbound assessment to show us where work is getting stuck and what outcome matters most.
Discover Your AI Opportunities