How to Do AI Employee Recognition Without Sounding Fake

Traditional employee recognition has four problems.

1. There isn't enough of it.
2. It arrives too late.
3. It's generic; doesn't really explain the value of the work, or what was hard about it.
4. It's time consuming to produce.

The result: on the infrequent occasions recognition happens, it calls out work that wasn't exceptional, or it prioritises the organisation's most visible teams at the expense of others.

When this happens, recognition's most important function, elevating the behaviour the organisation wants more of, stops working. Standards flatline or decline, as do satisfaction and retention.

As we outlined in What Is a Company Story Engine?, recognition has the same problem all content generation does: reviewing information to discover ideas, drafting the recognition, formatting and preparing it, approving it, and uploading it into the channel where you'll send it.

Simple AI systems go a long way here.

How to do employee recognition with AI

An AI employee recognition system produces specific feedback, based on a team or individual's deliverables, for a manager to approve.

It has four components:

• Work monitoring (exists today). Pull requests, tickets, posts, ad variants, support threads, call transcripts, CRM entries. The initiatives team members pursue in the course of their work.
• Org structure (exists today). Permissioning and user accounts, so you know who did the work, what their role is, and who they report to.
• Objectives and metrics context (exists today). OKRs, targets, roadmaps; most organisations have these.
• A system for tying them together (doesn't exist yet). This piece pulls it all together, decides what's noteworthy, and prepares it for review.

Depending on how ambitious you want to be, you can either make the recognition fully automatic, or route it to a manager to provide additional context and approve.

Now, instead of remembering to provide recognition and looking for ideas, your system is prompting your managers at relevant times.

How to get started

The easiest place to start is your HR system. Everything relevant is already there: job title, team, competencies, start date, end of probation date, years of service. Stories you can tell with this information include:

• passing probation
• years of service milestones
• promotions

For example:

That alone is enough to get started. From there, start pulling in more context, either automatically, or with a manager filling in the gaps:

You don't need AI to do all of the work. It's enough to prompt your managers when meaningful work or milestones are delivered, and use that to greatly increase the consistency and frequency of recognition.

How to avoid producing bad recognition with AI

One comment we hear a lot is AI can't do good recognition because it's depersonalised, doesn't come from a human, seems low-effort, etc.

Have you seen what the HR industry publishes as the standard of recognition? Here are two examples from the biggest names in the business:

Your outstanding performance and unparalleled work ethic are going to take you far in this company.

— Indeed

Your excellent performance is an inspiration to all. Keep up the great work!

— O.C. Tanner

Talk about low effort. The first is from a sample recognition letter published by Indeed, the largest job site in the world. The second sits on a menu of sixty ways to say thank you, published by O.C. Tanner, one of the largest employee-recognition companies in the world.

How to tell if your recognition is good: Use the Swap Test

The Swap Test: if a piece of recognition could be addressed to anyone in your company and still read correctly, it's bad recognition.

Take the proposed message and re-address it to your newest hire, your best employee, someone on a performance plan. If it could apply to anyone, it's a waste of time.

Recognition is a feedback loop, and the best recognition has four properties:

• It could only apply to one person(/team)
• It uses a situation-action-result type of narrative to explain why the recognition is important
• It happened recently enough that the person(/team) is still thinking about it
• It reinforces both the work that individual did, and clarifies the standard for others who observe the recognition being awarded

When you are specific enough - AI or otherwise - you will get great recognition.

What the output should look like

The pattern: what happened, what the result was, and why it's noteworthy.

Read the ✓ column. No vague superlatives; lots of specifics about what exactly somebody contributed.

"Using AI for employee recognition is lazy and deceptive"

This is false. AI employee recognition is simply about systematising recognition so that it happens more consistently, more accurately, and at a higher quality level than is possible with human-led recognition systems alone.

It is not either-or. Use AI to systematise the 80-90% of recognisable moments humans don't have time for; allow humans to come in over the top on the 10-20% of really big moments.

You have probably seen this scenario:

Steady contributor of ten years finally gets the limelight, and comes out of their shell. Nothing was done to them. They were not manipulated. A long injustice was corrected, they re-evaluated how much the organisation values and trusts them, and they found the confidence to stretch further.

Recognition tells people where they stand. A generic "great job" changes nothing, because it says nothing. Being specific shows that the organisation really understands their contribution.

Let us help you

If you want recognition running off your team's real work, and you'd rather not build that fourth component yourself, get in touch. I'll embed with your team, wire Fireside into the systems where the work happens, and get drafts landing in front of your managers so all they have to do is read them and hit "send".