Defined jobs
AI agents with a defined job and a handoff
An AI agent at New Era AI is software with a bounded job: answer a class of questions, qualify a lead, update a CRM, or run a spoken call. It is not a personality bolted onto the website with no owner.
In plain terms
Voice agents, chat agents and CRM agents share the same design rules: a scripted outcome, a data boundary, and a handoff to a person when the task needs judgement you have not encoded.
Echo One is the calling platform when the job is speech. Nexus One is the CRM product when the job is pipeline. Custom agents are Execute work on top of consulting.
The design test for any agent is the same: could a new member of staff follow the same script, know when to stop, and leave a record behind. If the answer is no, the agent is not ready to face a customer.
The problem
An agent without a job description becomes a novelty chat window.
Customer conversations and internal questions wait for whoever is free. Response times drift, answers vary, and the CRM never hears about it.
The wrong fix is an unbounded bot. The right fix is a job: what it may say, what it may book, and when it must stop.
How this fits
Discover names the job. Roadmap says whether an agent is this quarter. Execute builds it (voice, chat or CRM). Support reviews transcripts and handoff rates as the process changes.
Kinds of agent we deploy
- 01
Voice agents
Spoken inbound and outbound calls through Echo One, with transcripts into the CRM. Covered in depth on AI voice agents.
- 02
Chat and messaging agents
Web, WhatsApp or email front doors that capture intent and pass complex cases with context.
- 03
CRM agents
Score, follow up and move stages inside Nexus One or the CRM you already run.
- 04
Handoff rules
Named conditions for transferring to a person: complaint, regulated advice, or anything you have not scripted.
- 05
Internal agents
Staff-facing helpers for process answers and draft follow-ups, still bound to your data rules.
Message or call through an agent
- 01
Arrive
Call, chat or WhatsApp hits the agent with a known identity where you have one.
- 02
Do the job
It follows the script: answer, capture, qualify or book within the rules.
- 03
Write
Outcome lands in the CRM. If it cannot, the design is unfinished.
- 04
Handoff
A person takes over with the transcript when the rules fire. The agent does not improvise a regulated answer.
Where this actually gets used
Front door for common questions
Opening hours, service area, “can you do X”: answered consistently, with a person for anything else.
Qualification before a salesperson spends an hour
The agent asks the questions your team already asks, then books or logs a no. It does not invent a pitch.
After-hours coverage without a night shift
Voice or messaging takes the enquiry, writes the record, and queues the morning callback.
Internal “how do we do this?”
A staff agent that looks up the process you documented, rather than pinging the one person who remembers.
Where agents earn their keep
- Customer service teams buried in repeat questions
- Sales teams that lose evenings and weekends
- Ops teams that are the walking manual
Related products
Read next
Questions
AI agents, answered.
Is an AI agent the same as a chatbot widget?
A widget with no CRM write and no handoff is not what we ship. An agent has a job, a record and a stop rule.
Voice or chat first?
Whichever channel currently drops the most work. Voice is often missed calls. Chat is often the website. Discover picks.
Can an agent give professional advice?
No. It should not replace regulated or professional judgement. Those cases hand off.
Who builds Echo One agents?
Echo One is developed and operated by New Era AI Ltd. Custom jobs on top of it are still Execute work.
Will staff lose their jobs?
The design is coverage and consistency for repeatable work, with people on exceptions. We do not sell headcount claims.
How do we start?
A free assessment, then a bounded job on the roadmap. We do not start with an unbounded “AI employee”.
How do we keep an agent from saying the wrong thing?
By bounding it: a defined script, a data boundary, and named conditions that hand to a person. Support reviews transcripts so drift is caught early. An agent that is unsure should escalate, not guess.
Next step
Start with the free AI Opportunity Assessment.
A short, no-obligation conversation about where enquiries, hours and revenue leak today. You do not have to pick a tier to have it, and what comes out of it feeds Discover, so the first paid day starts from evidence rather than a blank sheet.