<figure class="post-ascii-fig">
<pre class="post-ascii" aria-label="Density-shaded ASCII illustration of two figures, one looking up at a larger tier above">
.                                .=#%##*                                                            
.                               -%%@@@@@%                                                     ..... 
..                              #@@@@@%#=                                                   ...     
.....                            *@@@%*=.                     .                            .......  
......         .                 -#%####*.                 +%@@%%=                   .              
.......                           #%##**=   -==:.         =@@@@@#+:                             ....
.........       ...:             :%%%%%#  -+=##:++=++==.  *##%#*#+                            ......
...........   ..:-=++*******++==+@%@@%+  ##*=*=+###@%%%*++%%%%%*+   =*                           ...
:..........   =*#%*#%%%%%%%%%%#*=%**#=#**.   *+   +@@@@@%#@@@@@#: :=*+                              
........... +#%%=  ..     .@@@#**%+=--##%+  :*+  =@@@@@@%#%%#+#   +++.                             .
. ...... .+%@#.        ....%%%#*@@==*+@@@#+:+*= :@@@@@@@@####*%: :%#                                
:.:.:--:+%@*..          . :%**#@@@@+++@ +@@%**= @@@@@@@@@%%%%%##*#+                               ..
--+=+**+#%=:.:..........:.=@@@@@@@@@@@@  @##**-%@@@@@@@@%%%*-%%##+                                  
:=+===%**+*=:::.....:.:::.=@@@@@@@@@@@    +%#*#@@@@@@@%%%+   =#*.                               ..  
=+***++%%#*=:::--:::::.:..=@@@@@@@@@-  .....%%@@@@@@@@@+.                          ..               
**+*+**--*##*=--=::-..::..=##%@@@@@-...--.+@%%%#%@%@@@=      .. .             .......               
+++++==+*==#+-==-=---::-:#@@@%###+:..:.:-#@@@@@@@@@@%. .  .....         ....... ........ . .........
+=++++#%##+-:=--==----=#@@@@@@@@%=:::---=@@@@@@@@@@@@%+:..........     .. ..:=-.       ...... ....:.
#*+==+*%%*=:---::::::#@@@@@@@@@@===-+++++*@@@@@@@@@@@@-:. ...:...........::-:-:.....      ..::::::..
%%##++**=+==:--=-----@@@@@@@@@%++####%##%%@@@@@@@@@@**++=-:  ......-::=-:::-++==-:.............:::::
@@@%#****+**==*===---@@@@@@@@@@@@@##%##%%@@@@@@@*%@%##%###*=: :....:==+=-===##=+=---=--. ......:::::
%@@@%###**++-+%*%#***@@@@@@@@@@@@@@@%***@@@@%@@:....::##*##**:.:.::-++#*+*-=##*++==++=+-:..:.....::.
%%@#+=+++=====**+---@@@@@@@@@@@@@@@@@@@#=+#**=:.....+#***##=:...-----==::=--==+++==+===-=++-==-....:
++===-==---:::-::--#@@@@@@@@%%@@@@@@@@@@@%*+:::::.:+*###*=......::-:--:-------=+++==-==-=-::-::::...
-::::::::::..:.:::-@@@@@@@@=--=-==+@@@@@@@@::::::=**##=:....::::.::::.::-=-::::::.:::::::::-::::::--
-:::----------==::@@@@@@@%::.....+@@@@@@@@=:. :.+*#=. . ....................:::.:::::::::::::::-----
... .            *@@@@@@:      :@@@@@@@+.                .      . ...   ..............:..:....::::::
.........      .%@@@@@@      .@@@@@@@.                                       ....    ....        ...
              +@@@@@@:     .*@@@@@+                                                       .         
........... .*@@@@@-     .*#*#%=-:          .  .                                          ..... ....
...:........#@@@@*....   . .. ::.  ..             ..       . ..... .    ... ... ............ .......
</pre>
<figcaption class="post-ascii__caption">fig 0. <code>tier(self)</code></figcaption>
</figure>

In Darwin we use a tier system to self-assess how each person on the team uses AI. Three levels — we don't even count the people who don't use it.

<pre class="ascii-art" aria-label="The three AI tiers, narrowing toward tier 1">
                  <span class="accent">/\</span>
                 <span class="accent">/  \</span>
                <span class="accent">/ 1  \</span>           <span class="label">"what should I do?"</span>
               <span class="accent">/------\</span>           <span class="muted">AI proposes; you ride</span>
              /        \
             /    2     \         <span class="label">"do this task"</span>
            /------------\         <span class="muted">AI executes; you direct</span>
           /              \
          /      3         \      <span class="label">"translate this email"</span>
         /------------------\      <span class="muted">AI as a fancier Google</span>

       most people  -> tier 3
       your team    -> tier 2 (minimum)
       your edge    -> tier 1 (rare)
</pre>


**Tier 3** is using AI like Google. Drafting an email. Looking up a recipe. Translating a paragraph. Most of these things could already be done with pre-GPT software, so you're not really getting anything new. If someone says *"yeah, I use AI all day to write emails,"* that's tier 3.

**Tier 2** is using AI to do *tasks*. Not just drafting an email — running every email through Claude for triage. Not just writing a marketing brief — having AI do deep research first, build the landing page code, and check the copy. Compressing steps that used to take hours. Most people on a serious team should be at least tier 2.

**Tier 1** is the hard one. Tier 1 means asking AI to *propose things you wouldn't have thought of yourself*. Not *"help me build this fundraising deck"* but *"should I be raising at all? Look at our metrics, look at my situation, tell me what to do."* Not *"write copy for this campaign"* but *"given everything we know about our customer, what campaign should we run?"*

The scary thing about tier 1 is that today's models still hallucinate. You can be wrong. But as the models keep improving, tier 1 will start to dominate tier 2 and tier 3 — because tier 1 doesn't cap the AI at the human's prior judgment.

We don't know yet whether you want a whole team of tier 1 people or a mix. Probably a mix. But you want zero tier 3 and zero non-users. That's a baseline.

Self-assess honestly. The trap is that tier 3 people often think they're tier 1, because they've used ChatGPT a lot. The number of hours doesn't matter. What matters is the kind of question you're asking it.