Using the solution found in this thread:

I'm able to parse the top level elements in the following schema (Google DocumentAI response):

ARRAY<
  STRUCT<
    mentionText STRING,
    id INTEGER,
    type STRING,
    confidence NUMERIC,
    properties ARRAY<
      STRUCT<
        normalizedValue STRUCT<
          text STRING
        >,
        mentionText STRING,
        id INTEGER,
        type STRING,
        confidence NUMERIC
      >
    >
  >
>

with this query:

EXECUTE IMMEDIATE '''SELECT ''' || (SELECT STRING_AGG(DISTINCT "MAX(IF(e.type = '" || e.type || "', e.mentionText, NULL)) AS " || e.type),FROM dataset.table, UNNEST(entities) as e) || '''FROM dataset.table, UNNEST(entities) as e'''

to transform this JSON:

{
    "entities":
    [
        {
            "id": "0",
            "mentionText": "John",
            "type": "first_name"
        },
        {
            "id": "1",
            "mentionText": "Smith",
            "type": "last_name"
        },
        {
            "id": "2",
            "properties":
            [
                {
                    "id": "3",
                    "mentionText": "45.00",
                    "type": "amount"
                },
                {
                    "id": "4",
                    "mentionText": "€",
                    "type": "currency"
                }
            ],
           "type": "transactions"
        },
        {
            "id": "5",
            "properties":
            [
                {
                    "id": "6",
                    "mentionText": "128.00",
                    "type": "amount"
                },
                {
                    "id": "7",
                    "mentionText": "€",
                    "type": "currency"
                }
            ],
           "type": "transactions"
        },
    ]
}

into this: