How to practice sales conversations with an AI agent in Tigy
Configure a fictional character and review practice manually with your team.
- Author
- Tigy AI team
- Published
- Updated
Sales practice with a voice agent simulates a conversation to rehearse discovery, listening and objection handling. In Tigy AI, a prompt agent can represent a fictional character through configured instructions. Define the scenario, facts and boundaries before testing. This guide uses human review and observable moments rather than assuming automatic scoring, model training or proven sales improvement.
Write the scenario
Choose a fictional buyer for a simple service with a goal, constraint and main question. For example, a small store seeking to organize order inquiries. Exclude real customer names, phone numbers and histories from the shared prompt.
Instruct the character
Explain that the agent participates in rehearsal, answers one question at a time and preserves scenario facts. Specify what it knows and what it cannot invent. Attach no real write, transfer or lookup tools to this exercise.
Test in the editor
Save the prompt agent and use available text or voice testing. Try an introduction, open question and problem confirmation. If character facts change, adjust one instruction and repeat that segment to verify correction.
Review with human criteria
Staff can mark whether a question preceded the proposal, understanding was confirmed and next steps were explained. Separate observation from interpretation and cite practice utterances. The character is not demand evidence and review is not a native automatic score.
Vary one difficulty
Once stable, change one constraint such as available time or missing information. Preserve the rest for comparison. This procedure uses prompt configuration and manual review rather than assuming native coaching or simulation.
How do you write a useful sales-practice character?
Define the exercise goal, fictional buyer context, need, known information and coherent objection. Instruct the character not to invent contradictory facts or leave its role to evaluate the seller. Use fictional data instead of reproducing a real customer profile.
For example, practice discovering needs before presenting an offer. The character says they need shorter service times but explains the operation only when asked. Staff review whether the seller clarified tasks and constraints; the simulation alone does not demonstrate sales conversion.
Choose one sales skill per exercise
Sales practice with a fictional character works better when the team knows what it is exercising. Discovering needs, explaining value, handling an objection, and agreeing on a next step are related skills, but need not carry equal weight in every session. Choose an observable objective and build a scenario that allows it to be practiced. This prevents long conversations where the salesperson talks extensively without receiving useful feedback on the selected skill.
At a fictional service company, the goal might be investigating why a prospect wants to change suppliers. The character knows the problem and answers questions but does not reveal every detail in its opening statement. Review focuses on whether the salesperson asked about context, impact, and priority before presenting the offer. Success need not mean a sale: it may mean accurately understanding the need and proposing an appropriate next step.
For another session, the objective might be explaining a proposal without jargon. Here the character asks questions when the salesperson uses unfamiliar terms. The reviewer checks whether the explanation became understandable and preserved the offer's actual conditions. Do not combine this exercise with an endless stream of objections, because that changes difficulty and makes assessment less clear.
Tigy lets the team configure an instruction-based agent to play this character. The practice described here requires team planning and review; it does not assume a validated automatic scoring system. Before opening a test, write the skill, situation, expected character behavior, and signals the reviewer will observe in the salesperson's performance. A defined purpose makes both the conversation and the later feedback more concrete.
Give the character consistent facts and role boundaries
A useful character needs enough context to answer consistently. Define role, fictional company situation, problem, priority, constraints, and knowledge of the offer. An extensive biography is unnecessary. Facts should affect the conversation: if the person cannot decide alone, they should explain how decisions happen when asked. If budget is limited, that constraint should remain consistent rather than change to make the exercise harder.
Consider a fictional prospect managing a small group of stores and receiving many calls about opening hours. Their problem is repetitive questions during busy periods. The character can describe the routine, identify who participates in a decision, and ask how the solution handles outdated information. These details let the salesperson investigate needs and explain conditions without forcing the agent to invent an entire business during the session.
Specify what the character should not do. It should not leave its role to teach the ideal response mid-conversation, assign a score, or pretend it signed a real contract. It may express interest and agree to a fictional next step according to the scenario, but the interaction remains practice. Do not attach production tools creating actual commercial records simply to make the session feel more realistic. Keep any actions within an appropriate test setup.
Use fictional information and check consistency before giving the exercise to staff. Ask direct questions about the problem, budget, and decision process. If the character contradicts its own brief, the salesperson's difficulty no longer reflects commercial skill; it reflects an unstable exercise. Correct the instructions and repeat the check. A believable character is useful because it responds coherently to the participant, not because it surprises the salesperson with arbitrary facts that were never part of the scenario.
Use objections that follow from the scenario
An objection helps practice when it represents a plausible concern for that character. “It is expensive” may mean insufficient budget, value not yet understood, comparison with another solution, or reluctance to make a commitment. The salesperson should have an opportunity to investigate before answering. A character rejecting every statement regardless of content turns the exercise into arbitrary resistance.
For the fictional store prospect, a possible objection is the fear of giving incorrect opening hours. That follows from the updating problem and allows discussion of information sources, maintenance, and service limits. Another concern may involve continuity when callers want a person. The salesperson should explain the offer's real capabilities and recognize what depends on configuration or staff. Promising any feature to satisfy the character should not count as a good answer.
Define how the character responds to clarification. If a concern has been addressed appropriately, it can acknowledge the explanation and move forward. If the response misses the concern, it can ask again about a specific aspect. This keeps difficulty tied to the objective. Avoid instructing the agent to always introduce another objection, because the salesperson would never have a way to demonstrate that the first one was addressed.
Include a case where the offer does not meet the need. Appropriate commercial behavior may be acknowledging the mismatch, explaining limits, and avoiding pressure for an unsuitable proposal. Practice should strengthen judgment and listening rather than only speed in countering objections. During review, ask the participant what they understood about the concern and why they chose their response. That reflection helps distinguish memorized phrasing from a contextual decision. It also shows whether the participant recognized when the next useful step was clarification rather than persuasion.
Review observable moments and give actionable corrections
After the conversation, select several moments representing the practiced skill. Feedback centered on “that was good” or “you lacked confidence” does not show the salesperson what to do differently. Identify the question asked, the information it revealed, and the missed opportunity. If discovery was the objective, check whether the participant asked about impact before recommending a solution.
Use a simple review sheet prepared by the team. It may examine whether the participant confirmed the need, explained the offer accurately, acknowledged a limitation, and agreed on a coherent next step. Interpret these criteria in context. A specific sentence is not mandatory, and the call should not become a recited checklist. Good performance may use different wording while achieving the same observable result.
Ask for a brief self-assessment first. The participant may identify where they assumed information, interrupted the character, or answered without clarifying an objection. Then compare that interpretation with conversational evidence. The reviewer should distinguish performance problems from scenario problems: if the character changed facts without reason, the session cannot fairly be compared with other exercises. Record that limitation and repair the scenario before drawing broad conclusions about the salesperson.
End with a correction that can be practiced in the next attempt. “Before presenting, ask about the effect on daily operations” is actionable. “Be more consultative” still requires interpretation. Choose one or two changes for repetition instead of listing every weakness at once. Review here is performed by people using explicit criteria; having a transcript or an AI character does not establish automatic assessment or validated commercial competence. Evidence should support the particular feedback being given rather than an unsupported overall judgment of the person's ability.
Increase difficulty without changing the assessment rule
A practice sequence may begin with a character that answers directly, then introduce incomplete information, interruptions, or a more specific concern. Difficulty should increase intentionally. If discovery remains the objective, keep discovery criteria while changing how the character presents its problem. This allows progress to be observed without comparing exercises demanding entirely different skills.
Repeat a scenario after feedback, then use a similar one to check transfer. A participant may memorize the right question for the first brief; a second situation reveals whether the principle was understood. For example, replace the store prospect with a service company experiencing repetitive calls while retaining the need to investigate impact and priority. Avoid introducing dozens of new conditions, because that makes it harder to establish what actually changed.
Retain the character-instruction version and each session's objective. When several people practice, use comparable difficulty and aligned reviewers. Small differences in agent behavior may affect the experience, so check consistency before using results for consequential decisions. An informal exercise should not be presented as a validated staff-selection or performance-assessment instrument. Its evidence supports feedback about the observed session and the explicitly chosen skill.
Maintain connection to actual service through approved offer examples, frequent questions, and known limitations. The exercise should encourage accurate promises and executable next steps. When a product or service changes, update character briefs so salespeople do not practice outdated terms. The expected benefit is a safe opportunity to converse, receive feedback, and repeat with purpose; usefulness depends on scenario design and human review quality. Keep practice tools and fictional data separate from real customer operations, especially if a participant explores creating a lead, scheduling a meeting, or sending a follow-up during the role-play.
