Interview process best practices

A good interview process is a short, structured sequence in which each stage tests one thing, every interviewer asks job-related questions from a shared plan, and the team decides together from written evidence. It respects candidates’ time and moves fast. In 2026 it also has to account for AI, both the tools candidates use and the tools you use. This guide walks through designing the loop, running interviews, handling technical rounds and deciding, with the steps to run each part in Dover’s ATS.

Step 1: Design your interview loop

Start from the job, not a template. List the three to six things someone must do well in the first year, then assign each to a stage so nobody repeats another interviewer’s questions.

A common loop for most roles:

Fewer rounds is usually better. Three to five stages covers most roles, and senior or founding roles may justify one more. For each interviewer, name the competencies they own. Four interviewers asking “tell me about yourself” is wasted time. See how to interview candidates for your startup for a walkthrough of each stage.

Build it in Dover. Open a job and go to its Interview Plan tab to add, edit and reorder stages. Every plan starts with Applied, Responded and Offer stages, and you add the rest:

• Single-part interview: one meeting, such as a recruiter screen or hiring manager call. Set its name, duration, default interviewers, meeting type, email templates and feedback form. If you add more than one interviewer, Dover rotates between them based on availability.
• Multipart interview: several interviews on one day, such as an onsite. Add each part as a sub-stage with its own interviewers and feedback form.
• Assessment / Other: a take-home, reference check or any stage that is not a meeting.

You can also show a stage to candidates with a short public description, so they see what to expect. See Setting up your interview plan. Customizable interview plans are on the free and Premium plans.

Step 2: Make every interview structured

In a structured interview, every candidate gets the same job-related questions, asked in the same order, and scored on the same scale against the same standard. The US Office of Personnel Management lists consistency and equal opportunity to share information among the benefits. Published research on selection methods ranks structured interviews among the strongest predictors of job performance (Sackett et al., 2022).

To make an interview structured:

1. Define the competencies from the job.
2. Write two or three questions per competency, mixing behavioral (“tell me about a time”) and situational (“what would you do if”) questions.
3. Write what weak, acceptable and strong answers look like.
4. Ask every candidate the same core questions, allowing neutral follow-ups.
5. Score each answer on the scale as you go.

For question ideas, see our list of questions to ask during an interview.

In Dover, interviewers can open View AI suggested questions on the interview page to get questions based on the candidate’s background, each with a Why ask? explanation, and add the ones they want to their rubric. Use them to supplement your core questions, not replace them.

Step 3: Prepare your interviewers

Most interview problems start before the call. Before anyone interviews:

• Assign each interviewer a competency and the questions that test it.
• Share the scorecard and the standard. Interviewers should know what a 1 and a 4 look like.
• Brief them on what not to ask. See how to interview candidates inclusively.
• Give them the resume and any notes a day ahead, and tell them how to describe the company and role.
• Set their interview preferences. In Dover, each interviewer goes to Account settings → Interviews & Notetaker, adds their name and role, which candidates see, and chooses how candidates book time with them.

Step 4: Schedule fast and keep candidates informed

Speed is part of the design. In a Dover analysis of 300 companies, the faster group scheduled each next step within 24 hours, kept interviewer calendars open and sent candidates a note on who they would meet and what would be covered (the data). Those are cheap habits to copy.

In Dover:

• Send a scheduling link. From the candidate’s profile, add a scheduling link to your email, choose the interviewer and length, and let the candidate pick a time. You can use a Dover scheduling link or a third-party one such as Calendly.
• Place the meeting yourself. On Premium, Schedule Interview shows interviewers’ availability and creates calendar events for the candidate and the whole panel.
• Track outside bookings. Use Mark as scheduled for interviews booked outside Dover.
• Show candidates where they are. The candidate portal gives each candidate a status page with the stages you share.

See Scheduling candidates. Reply within a couple of business days after each stage, including rejections.

Step 5: Run the interview and capture evidence

During the interview, ask the planned questions, listen for specifics and take notes of what the candidate said and did, not your interpretation. Leave time at the end for their questions.

In Dover, interviewers open the candidate’s profile, click Add feedback and fill in the stage’s feedback form as they go. The page shows the resume beside the form. Two optional aids:

• AI Notetaker (Premium): joins interviews scheduled through Dover on Google Meet, Zoom and, with Microsoft Calendar, Teams. It records and transcribes the conversation and adds AI notes under yours. See Dover AI Notetaker.
• AI feedback forms: after the interview, paste in your notes or the transcript and Dover fills in each question. Submitting overwrites answers you already entered, so review the result.

Whatever you use, you still own the scores. See Interviewing candidates.

Step 6: Score with scorecards and rating scales

A scorecard lists the competencies for a stage and a rating scale with anchors. Use a scale of 4 or 5 points, define each point in words and add a notes field for evidence. A 4-point scale has no midpoint, so interviewers must lean one way. Interviewers should fill it in during or immediately after the interview and before they talk to colleagues, so the first opinion in the room does not sway the rest. Our interview scorecard guide walks through building one.

Build it in Dover. Create a form in Settings → Feedback Forms, or from a stage on the interview plan. Start from one of Dover’s role-based templates or build your own, then assign the form to the stage. Interviewers score on the default 1 to 4 scale, and on Premium you can create custom rating scales. Free includes a limited number of forms and Premium is unlimited. To see everyone’s feedback on a job, a member needs the View all feedback permission, which the hiring manager and recruiter get. See Interview feedback forms.

Step 7: Run technical interviews in the age of AI

Technical interviews are where AI has changed the most. Candidates can use assistants in real time, and the tools they will use on the job now write much of the code. Your process should test what the work actually requires.

Decide your AI policy for each round

There is no single standard yet. Trade reports and policy trackers say many companies still ban AI in live coding rounds, while a growing number allow it or build rounds around it, and some have gone back to in-person interviews to limit cheating. Policies change often and the reporting is secondhand, so treat any tracker as a starting point (one example). Choose a policy that fits the job and apply it to everyone.

Whichever you choose, put it in the invite, so candidates are not guessing, and keep one human accountable for the result.

Test what the job needs

• Prefer realistic work to puzzles. Debugging, reading and reviewing code, extending a small codebase, designing a system and explaining trade-offs show more than a memorized algorithm.
• If AI is allowed, test judgment. Give a task where the first AI answer is plausible but flawed, and see whether the candidate catches it. Change the requirements halfway through to see who is steering.
• Ask them to explain. Have the candidate talk through why they accepted or rejected a suggestion, how they tested it and what they would do next. Rehearsed or generated answers fall apart under follow-ups.
• Keep a short unaided check only for the skills the job needs without tools.
• Keep tasks short and fair. Two hours is a sensible ceiling for a work sample. Say how long you expect it to take, offer a live alternative to unpaid homework and score every submission with the same rubric.

Scorecard for an AI-assisted technical round

TECHNICAL INTERVIEW SCORECARD (score each 1 to 4, with evidence)

Problem framing: Did they clarify the goal and constraints before coding?
Approach and design: Is the plan sound, and do they explain trade-offs?
Working with AI tools (if allowed): Do they direct the tool well, check its output and reject bad suggestions?
Verification: Do they test, debug and reason about correctness and edge cases?
Code quality and reading: Can they read unfamiliar code and improve it?
Communication: Can they explain what they did and why, and take feedback?
Security and privacy judgment: Do they avoid pasting sensitive data into tools?

Overall: strong yes / yes / no / strong no
Evidence: [what the candidate said and did]

Protect the integrity of the round

• Check the candidate is real. Identity fraud and deepfakes are a documented problem in remote hiring, so run the same checks in technical rounds that you run in the screen, and look for the same person across rounds. See how to run a recruiter screen.
• Use live pairing and follow-ups instead of heavy monitoring. Our guide to preventing cheating on technical interviews covers proportionate options.
• Consider an in-person or controlled final round for sensitive roles.

For a complete engineering example with role-by-role questions, see how to hire engineers.

Step 8: Use AI responsibly in your own process

You may use AI as well, so set rules for interviewers.

• Tell candidates before you record or transcribe, and get their agreement. Some US states require consent from everyone on a call (Justia 50-state survey), and an interview is a poor place to surprise anyone. Let candidates opt out without penalty. In Dover, the notetaker joins as a visible participant called Dover Notetaker, the host can deny it entry, and recordings stay in Dover rather than going to candidates.
• Review AI notes before you rely on them. Summaries can be wrong or leave things out, so check them against what you remember.
• Do not let AI make the decision. Using AI to score or rank candidates can trigger audit and notice rules in some places. If you hire in New York City, read our guide to complying with NYC Local Law 144. Keep a person accountable for every decision.
• Protect candidate data. Share recordings and notes only with people involved in the decision.

Step 9: Run the debrief and decide

1. Collect written scores before the meeting. Nobody reads others’ feedback first.
2. Start with a round of independent scores, with the most junior interviewer speaking first.
3. Discuss disagreements by pointing to evidence from the interview, not general impressions.
4. Apply the decision rule you set in advance. For example: hire if no competency scores below 3 and the hiring manager and at least one other interviewer are a strong yes. Without a rule, the loudest voice wins.
5. Record the decision and the reasons.
6. Tell the candidate quickly, ideally within a couple of business days.

In Dover, turn on Debrief for an interview stage to tell your recruiter to schedule a debrief meeting after it, and choose its length. Debriefs are a Premium feature. After the decision, move the candidate to the next stage or reject them with a saved email template, and keep the reason in their profile.

Step 10: Protect the candidate experience and reduce bias

Candidates judge your company by how you run interviews, and consistent structure is also your best defense against bias.

• Tell candidates the stages, who they will meet and how long each takes before the process starts, and share each interviewer’s name and role.
• Offer flexible times and accommodations, and make scheduling easy.
• Use the same questions and scale for every candidate, score before discussing and decide from evidence in the notes.
• Replace “culture fit” with specific values and behaviors you can ask about.
• Use a mixed panel where possible, and give every interviewer the same training.
• Audit outcomes occasionally: where do different groups drop out of your funnel?

See interviewing candidates inclusively for what you may and may not ask, de-biasing job descriptions, and our candidate experience guide.

Interview plan template

Role: [title]            Hiring manager: [name]
Must-have competencies: [1] [2] [3] [4]

Stage 1: Phone screen      [20-30 min]  Interviewer: [name]  Tests: [competency]
Stage 2: Manager chat      [45 min]     Interviewer: [name]  Tests: [competency]
Stage 3: Skills exercise   [1-2 hrs]    Reviewer: [name]     Tests: [competency]   AI policy: [not allowed / allowed / required]
Stage 4: Final loop        [2-4 hrs]    Interviewers: [names, one competency each]
Debrief:                   [date]       Decision rule: [e.g. no score below 3 and manager is a strong yes]
Candidate told within:     [2 business days] of each stage

Interview feedback form template

Candidate: [name]   Role: [title]   Interviewer: [name]   Stage: [stage]

Competency: [name]
Question asked: [question]
Evidence (what the candidate said and did): [notes]
Score: 1 weak / 2 mixed / 3 good / 4 strong

[repeat for each competency you own]

Concerns to probe in a later round: [notes]
Overall recommendation: strong yes / yes / no / strong no
Submitted before reading others' feedback: yes / no

Frequently asked questions

How many interview rounds should I have?

Three to five stages is enough for most roles: a screen, a hiring manager conversation, a skills exercise and a final loop. Every added round should test something no earlier round tested.

What is a structured interview?

Every candidate for a role gets the same job-related questions, scored on the same scale against the same standards. The US Office of Personnel Management describes it this way and recommends it for fair, consistent assessment.

Should candidates be allowed to use AI in technical interviews?

Decide for each round and say so in the invite. Many teams still ban outside help in live coding screens, while a growing number allow or require AI in a realistic task and score how the candidate directs, checks and explains it. Keep a short unaided check only for skills the job needs without tools.

Should you use take-home assignments?

Only if they are short, relevant to the job and followed by a live conversation about the work. Tell candidates the time you expect and what you will evaluate.

Can interviewers use an AI notetaker?

Yes, if you tell the candidate first and get their agreement. Some places require consent from everyone on a recording, and some require notice when AI is used in hiring. Let candidates opt out without penalty, and keep a person accountable for every score and decision.

How do I set up an interview process in Dover?

Open a job, go to its Interview Plan tab, add each stage, choose its interviewers, length, email templates and a feedback form, then send candidates a scheduling link from their profile. Interviewers score candidates on the feedback form after each interview.